So far in this series I have argued that every asset is a claim on a functioning system, and that even money itself is a claim on confidence. This essay turns to the mirror image of an asset: a debt. Because for every loan that is someone’s burden, there is someone else for whom that same loan is an asset — and when the AI transition makes debts impossible to repay at scale, the question is not whether the loss happens. It is who absorbs it. The loss never simply vanishes. It moves. And following where it moves tells you a great deal about how this transition will actually be felt.

Let me start with the nature of debt itself, then look at the clearest current example, and then at the uncomfortable general principle underneath.

Every Debt Is a Bet That Tomorrow Resembles Today

A loan is, at its heart, a prediction. When you borrow, you are betting that your future self will have the income to repay. When a lender lends, they are betting the same thing about you. Both sides of every loan rest on a shared assumption so basic it is almost never spoken aloud: that the future will resemble the present closely enough for the repayment to be possible. The student borrows against the career the degree is supposed to produce. The homebuyer borrows against decades of expected earnings. The business borrows against the customers it expects to keep. Debt is how we move expected future prosperity into the present and spend it now.

This works beautifully as long as the future cooperates. And for most of modern history, instable times, it has — incomes have been roughly predictable, careers roughly durable, the ground roughly stable beneath the bet. Debt has been one of the great engines of human advancement precisely because the assumption underneath it usually held.

But notice what the entire system depends on: continuity. A loan is a bet that tomorrow will resemble today. And the AI transition is, more than anything else this series has examined, a force that threatens to make tomorrow not resemble today — to break the continuity that every debt silently assumes. When the career the student borrowed against no longer exists, when the earnings the homebuyer counted on erode, when the customers the business expected disappear, the bet fails. Not because anyone was reckless, but because the future stopped cooperating. And then the loss has to go somewhere.

The Clearest Example: $1.7 Trillion Borrowed Against a Vanishing Premium

Look at student debt, because it is the purest illustration of a debt that was a bet on the future — and of how that bet is now straining.

Americans owe roughly $1.7 trillion in federal student loans, spread across about 43 million people. Every dollar of it was borrowed on a specific premise: that a degree purchases a wage premium — that the educated worker would earn enough more, over a career, to justify the debt and then some. For generations, broadly, that premise held. The bet was reasonable.

Now lay the AI transition over it. The piece of this platform’s argument that matters most here is that AI is reaching first and fastest into exactly the kind of work many of these degrees were meant to lead to — the entry-level white-collar roles, the junior analyst and junior coder and junior associate positions, the knowledge work that was supposed to be the safe, educated, debt-justifying path. If the wage premium that the entire $1.7 trillion was borrowed against erodes — if the degree no longer reliably purchases the career it promised— then a vast number of these loans become bets on a future that is quietly being cancelled. The borrower did everything they were told. They invested in themselves, took on the debt the system encouraged, and pursued the credential. And the return on that investment is being undercut by a force that arrived after they signed.

The strain is already visible, before AI has even fully arrived. Heading into 2026, roughly one in four federal student loan borrowers — around 12 million people — are already delinquent or in default. The income-driven repayment structure that had been cushioning many of them was struck down, payments resumed for millions who had been shielded, and wage garnishment for defaulted borrowers is restarting. Experts in the field are openly using the language of a “default cliff” and warning of historic default rates ahead. This is the current reality before the full force of AI displacement reaches the cohort. It is a system already under stress, with the larger stress still to come.

And here is the question that makes it a Human Transition story rather than a personal-finance story: when those millions cannot pay — when the bet fails at scale — who eats theloss?

Following the Loss

This is the part almost no one traces honestly, and it is where the real structure of the transition reveals itself. When a debt goes bad, the loss does not evaporate. It lands on someone. Follow it.

In the case of student debt, the lender is, overwhelmingly, the federal government — which means the public. So when these loans fail at scale, the loss flows to the taxpayer, one way or another. It can happen explicitly, through forgiveness or write-offs, where the cost is absorbed openly and the political fight is about who deserved it. Or it can happen implicitly, through default, garnishment, and a generation locked out of forming households and spending and building the lives that drive an economy. Either way, the public absorbs it —either as a fiscal cost or as the macroeconomic drag of millions of impaired financial lives. There is no version where the loss simply disappears. There is only the choice of which form it takes and who is made to carry it.

Now widen the lens, because student debt is just the clearest case of a universal mechanism. Every category of debt has someone on the other side who is counting on repayment, and the AI transition threatens repayment across all of them at once. Mortgages that fail land on banks, and through them on depositors and the financial system — the subject of the next two essays. Corporate debt that fails lands on bond holders, pension funds, and the retirement savings of ordinary people who never knew they were exposed. Consumer debt that fails lands on banks and ultimately on everyone who uses the financial system. The web of debt is the web of the economy, and the defining feature of a web is that you cannot pull hard on one strand without the whole thing moving.

This is why “AI will displace some workers” is such a radical understatement of the stakes. The displaced worker is not an isolated event. That worker is also a borrower, and their default is someone else’s loss, and that someone is often an institution holding the savings of thousands of other people. The transition does not just hit the person who loses the income. It propagates outward through the debt that connected them to everyone else — which is to say, through the entire financial system.

The Honest Part: There Is No Clean Way to Eat This

As I promised at the start of this series, I will tell you where I stand and then admit where my own instincts run out of answers.

My instincts are toward personal responsibility and free markets. In a normal world, I believe people who borrow should repay, that debts freely taken on are obligations, and that a system which casually forgives debt teaches everyone to borrow recklessly and corrodes the trust that credit depends on. I do not think that instinct is wrong. But I also cannot honestly apply it without flinching to a situation where millions of people made a reasonable bet — invest in your education, the system told them, this is the responsible path — and then had the ground moved underneath them by a force they did not create and could not have foreseen. There is something genuinely unjust about telling a person they should simply bear the full consequences of a bet that was rational when they made it and was broken by a technological shift that benefits the very economy that encouraged them to borrow.

And yet the alternative is not clean either. Mass debt forgiveness, whatever its moral appeal in individual cases, has real costs and real unfairness of its own: it asks those who did not borrow, or who already paid, to absorb the losses of those who did; it can reward the institutions that profited from inflating the debt in the first place; and done carelessly it can feed exactly the currency debasement the last essay described, as governments simply print to paper over obligations that can no longer be met. The free-market reflex (let the defaults happen, let the losses fall where they fall) risks crushing a generation for a transition that was not their fault. The interventionist reflex (forgive, absorb, socialize the losses) risks unfairness, moral hazard, and monetary strain. Both roads have a real cost, and I am not going to pretend one of them is obviously right. What I am certain of is narrower: the loss is real, it is large, and it will be absorbed by someone — and the fight of the coming years will be over who.

That fight is worth watching closely, because the way a society chooses to distribute the losses of this transition is the truest test of what it actually is. The losses can be loaded onto the borrowers alone, or onto the public broadly, or onto the institutions that profited, or onto the future through debasement — and each choice reveals a different set of values about who matters and who is expendable. There is no neutral option. Choosing not to decide is itself a decision to let the loss fall wherever gravity and power take it, which historically means onto those least able to resist.

What to Take From This

Do not read this as an argument for or against any particular policy. Read it as a lens.

The lens is this: in a debt-saturated economy facing a continuity-breaking transition, watch the debt, because the debt is how the pain travels. When you hear that AI is displacing workers, do not stop at the worker. Ask who lent to that worker, who is counting on their repayment, whose savings sit on the other side of their default — and you will see the transition’s reach extend far beyond the people directly displaced, into institutions and savers and taxpayers who imagined themselves uninvolved. Almost no one is uninvolved. The web of debt connects the displaced to everyone, which is exactly why no one is as insulated as they feel.

And for yourself, the usable wisdom is old and unglamorous: in a time when the future is less likely to resemble the past, debt is more dangerous than it used to be, because debt is a bet on continuity and continuity is exactly what is in question. This is not a reason for panic,and it is certainly not financial advice. It is a reason to treat the assumption underneath every loan — tomorrow will resemble today — as the open question it has quietly become, and to build the kind of resilience that does not depend on that assumption holding. The household with less debt and more community is not richer on paper. It is simply less exposed to the one bet this whole era is calling into question.

The train is moving. And much of what we call wealth and obligation is just a vast set of promises about a future that is no longer promised.

Key Terms Explained

Debt as a bet on continuity: The core idea of this piece — every loan assumes the future will resemble the present closely enough for repayment to be possible. The AI transition specifically threatens that continuity, which is what makes debt newly dangerous.

The wage premium: The higher lifetime earnings a degree is supposed to provide, and the premise on which ~$1.7 trillion in student debt was borrowed. AI’s reach into entry-level white-collar work threatens exactly the premium that justified the borrowing.

Following the loss: The principle that a bad debt’s loss never vanishes — it moves to whoever is on the other side (taxpayers, banks, depositors, bondholders, pensioners). Tracing where it lands reveals how the transition propagates beyond the directly displaced.

The distribution question: Who absorbs the losses of mass default — borrowers, the public, profiting institutions, or the future via debasement. Each choice reflects a different set of values; there is no neutral option, and not choosing is itself a choice.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio, who writes from a free-market, personal-responsibility perspective and aims for independent analysis over political allegiance. Nothing here is financial advice.

This is the third essay in a sub-series on the financial system. Next: housing and mortgages — what happens when the foreclosures have no buyers.

Subscribe to The Human Transition on Substack to follow the whole series.

Sources and further reading:

• AEI (Preston Cooper) analysis of federal student loan data, late 2025 — ~12 million borrowers delinquent or in default

• NPR — “Federal student loans are changing: what to expect in 2026” (default cliff, garnishment resuming, SAVE struck down)

• The Century Foundation — “Trump’s Student Loan Delinquency Crisis, Unmasked,” 2026 (SAVE forbearance data, ~$400B, distress projections)

• EducationData.org — Student Loan Debt Statistics 2026 (~$1.7 trillion total, ~43 million borrowers)

• Congressional Budget Office — projections on income-driven repayment forgiveness costs

In the previous essay, I argued that every asset is a claim on a functioning system, and that the AI transition is straining the systems those claims depend on. I walked through property, stocks, bonds, and land. But there was one thread I deliberately left hanging, because it deserves an essay of its own. It is the deepest one, and the most disorienting, because it sits underneath all the others.

When I told you that a bond was only as safe as the currency it pays you in, I was pointing at something most people never stop to examine: the money itself. We measure the value of every asset in money — the house is “worth” so many dollars, the portfolio so many, the wage so many. Money is the ruler we hold up against everything else. But here is the question almost no one asks: what measures the ruler? What is the value of money itself measured against? And what happens if the ruler is not made of steel, but of rubber — quietly stretching and shrinking while we use it, confident, to measure everything else?

This essay is about the currency itself: what actually gives money its value, why that foundation is more fragile than we let ourselves believe, and what the AI transition specifically could do to the most fundamental system of all. I’ll keep my promise from the start of this series to disclose where I stand — free markets, sound money, limited government are my instincts — and my equally honest promise that I do not think those instincts contain a tidy solution to what follows. Let me show you the problem first.

What Actually Gives Money Its Value

Pick up a banknote and ask what makes it worth anything. It is a piece of paper, or increasingly just a number on a screen. It has no intrinsic value. You cannot eat it, build with it, or burn it for meaningful heat. And yet you can exchange it for food, shelter, a day of someone’s labor. Why?

The honest answer unsettles people when they first sit with it: modern money has value almost entirely because of shared confidence. We accept money in exchange for real things because we are confident that others will, in turn, accept it from us. The paper is backed not by gold, not by any physical commodity, but by a collective belief — a belief in the issuing government’s stability, in the broad acceptance of the currency, in the expectation that the dollar or euro or peso you hold today will still buy roughly the same amount tomorrow. Money is a confidence game in the most literal and non-pejorative sense: it works because, and only because, enough people have confidence in it.

This is not a flaw that a better policy could fix. It is the nature of money in the modern world. Every major currency on Earth is a “fiat” currency — valuable by decree and by collective agreement, not by any underlying physical backing. And this means the value of money rests on exactly the same kind of foundation as every other asset we examined: a functioning system, sustained by belief. The currency is not the bedrock beneath the assets. It is the deepest layer of the same conditional structure — and when confidence in iterodes, there is nothing underneath to catch you, because it was the bottom.

That is why this is the most important essay in the series. Everything else is measured against money. If money itself is unreliable, then every measurement you have made of your own safety was taken with a faulty instrument.

The Quiet Theft You Already Know

You do not need a hypothetical to understand how money loses value, because you have lived through a mild version of it. It is called inflation, and it is the most familiar form of the rubber ruler stretching.

When your money buys less this year than it did last year, the number in your account has not changed, but its value has quietly shrunk. Nobody took anything out of your account. Your savings were diminished invisibly, by the erosion of the unit itself. This is worth dwelling on because it reveals the mechanism that matters for everything that follows: you can be made poorer without losing a single dollar, simply by having the dollar made worthless. The theft leaves no fingerprints, because nothing was taken — the unit was just redefined underneath you.

Most of the time, in stable economies, this happens slowly enough to tolerate — a couple of percent a year, an ambient erosion we barely notice. But the same mechanism, under enough pressure, can run fast. History is full of currencies that lost most or all of their value in years, months, even weeks, when confidence broke. The people holding those currencies did not do anything wrong. They saved diligently in the money of their country, and the money betrayed them — not because it was stolen, but because the shared confidence that gave it value evaporated. Their bonds were repaid. Their bank balances were intact. And they were ruined anyway, because the unit those numbers were denominated in had become worthless.

Keep that mechanism in mind, because the AI transition presses on every one of the things that hold a currency’s value up.

How the Transition Presses on the Currency

So how, specifically, could the force this series has been documenting reach all the way down to the value of money itself? Through several channels at once, and they compound.

Consider what a government faces in a serious displacement scenario. Tax revenues fall, because the incomes and transactions that generate them are eroding. Simultaneously, the demands on the state explode — for income support, for safety nets, for the management of mass disruption. A government caught in that scissors has historically reached, again and again, for the same tool: it creates money. It spends what it does not have, financing the gap by effectively printing currency. And when the supply of money grows much faster than the supply of real goods and services, each unit of money is worth less. The rubber ruler stretches. This is not a fringe scenario; it is the most well-trodden path in monetary history, and the AI transition is precisely the kind of fiscal emergency that sends governments down it.

There is a second, subtler channel, and it is the one that should give pause even to optimists. Some of the most prominent voices in technology — I wrote about them recently— promise that AI will make everything cheaper, driving costs toward zero, producing a deflationary abundance where money stretches further and further. Set aside whether that is realistic. Notice instead that we are being told, simultaneously, two opposite stories about the future of money: that AI will make it more valuable (deflation, cheap goods) and that governments will be forced to debase it (inflation, money-printing to fund the crisis). The truth is that both pressures are real and they will fight each other, and the outcome is genuinely uncertain — which is itself the point. When credible people are predicting both significant inflation and significant deflation from the same underlying event, what they are really telling you is that the future value of money is unusually unpredictable. And unpredictability is itself a form of risk to anyone whose security is denominated in that money. You cannot plan around a measuring stick when you do not know which way it will stretch.

Even the King of Currencies Is Not Insulated

Now I want to address the comforting thought that many readers will be holding: this might happen to weak currencies — the pesos and the liras of the world — but surely not to the dollar, or the euro, the strong reserve currencies. Those are safe.

This is the most dangerous assumption of all, and the evidence is already quietly contradicting it.

It is true that strong reserve currencies, above all the US dollar, have enormous advantages — depth, liquidity, the trust of centuries, the fact that the world conducts most of its trade and holds most of its reserves in them. That “exorbitant privilege” is real and it is not vanishing overnight. But “not overnight” is not the same as “never,” and the direction of travel is unmistakable for anyone willing to look. The dollar’s share of global reserves has fallen below 57% — the lowest since the mid-1990s, down from over 70% at the start of the century. Central banks around the world have been quietly buying gold at record levels —not as an investment, but as a deliberate hedge against exactly the fiat-currency risk this essay describes, accumulating the one reserve asset that cannot be printed, frozen, or debased by anyone else’s decision. When the institutions that manage the world’s money are themselves steadily reducing their dependence on the dollar and reaching forsomething harder, they are telling you something. They are revealing, through action rather than words, that they do not regard even the king of currencies as unconditionally safe.

The historical precedent should sober anyone who thinks reserve status is permanent. The British pound was the world’s dominant reserve currency, holding around 90% of global reserves in the late 1940s. By the early 1970s that share had fallen to roughly 11%. The transition took decades and felt gradual — until, at the points of crisis, it was not gradual at all. Reserve currencies do not usually collapse in a day. They erode for years while everyone insists it cannot happen, and then they move suddenly when confidence finally breaks. The person who assumed the pound was eternal because it had always been dominant was not wrong for a long time. And then they were ruined all at once.

I am not predicting the fall of the dollar, and I want to be precise about that. The dollar may well retain its dominance for a long time yet; its advantages are genuine and its rivals have deep problems of their own. The point is narrower and more important: even the strongest currency is a claim on confidence, not a law of nature, and the AI transition is exactly thekind of multi-decade stress — fiscal, geopolitical, social — under which confidence is tested. To treat any currency as unconditionally safe is to make the same error as the landlord who thought the building was the asset, when the asset was really the solvency of the tenants. The dollar is not the bedrock. It is the most trusted of the conditional promises — which is not the same as unconditional, and the difference is everything.

What This Means, Honestly

Let me be careful here, because this is the territory where false prophets thrive, and I refuse to become one.

I am not telling you the dollar is about to collapse. I am not telling you to convert your savings into gold, or crypto, or canned goods. Predictions of imminent monetary doom have been a reliable way to sell newsletters and bad investments for fifty years, and most of them have been wrong. The honest position is not “the currency is doomed.” The honest position is this: the value of money is not the fixed point we treat it as. It is a variable — a claim on confidence that the AI transition will stress in ways pointing toward both inflation and deflation, with an uncertain result. And the unit in which you measure all your other safety is therefore itself uncertain.

That recognition changes how you should hold everything else. It means “I have X dollars saved” is a less complete statement of security than it feels, because the value of those dollars is itself in question. It means real resilience cannot be denominated entirely in any single currency, because to do so is to bet everything on one measuring stick staying rigid. It is part of why diversification — across currencies, across asset types, across the things that hold value through different kinds of stress — is not a luxury but a basic hedge against the unit itself failing. We will look hard, later in this series, at the specific things people flee to when they sense this — gold, which holds value across monetary regimes but which you cannot eat; crypto, which is sold as an escape but behaves in stress more like a risk asset than a refuge. Each deserves its own honest examination, and each will get one.

And here, as promised, is the admission that keeps me honest. What is the policy answer to the fragility of money in the AI age? My own instincts run toward sound money, monetary discipline, limited government spending — and I will tell you plainly that I do not see how those instincts survive contact with a genuine displacement crisis. A government watching its citizens lose their livelihoods at scale will not practice monetary restraint; it will print, because the political alternative is unbearable, and it will be reaching for the one lever that lets it respond to a catastrophe it cannot otherwise fund. I find that troubling. I do not have aclean solution to it. The interventionist’s answer — manage the money supply actively, deploy it deliberately to cushion the transition — runs straight into the history of every currency ruined by a government that believed it could print without consequence. Both roads have a cliff. I am not going to pretend I can see a safe path between them that the evidence does not support.

What I can tell you is the same thing this series keeps arriving at. If even the measuring stick is made of rubber, then security was never going to be found in accumulating units of it. It is found in resilience that does not depend on any single currency holding its value — in real things, in diversified holdings, and most of all in the web of people and capacities around you, which retains its worth no matter what happens to the number on the screen. The community that can feed and support itself does not need to ask what the currency is doing this week. That is not a complete answer to a monetary crisis. But it is the part of the answer that no government, and no act of printing, can take away from you.

The train is moving. And it is coming not just for your job and your assets, but for the ruler you use to measure all of them.

Key Terms Explained

Fiat currency: Money that has value by government decree and collective confidence rather than by backing from a physical commodity like gold. Every major modern currency is fiat — meaning its value rests on belief in a functioning system, the same conditional foundation as every other asset.

Unit of account: Money’s role as the common measuring stick for the value of everything else. The deeper danger this essay identifies: if the unit itself is unstable, every measurement of safety made in it is unreliable.

Debasement / inflation: The loss of a currency’s value, by which holders are made poorer without losing any nominal amount — the unit is simply redefined underneath them. Usually slow, but capable of running fast when confidence breaks.

Reserve currency: A currency held in large quantities by global central banks and used for international trade. Confers major advantages (“exorbitant privilege”) — but, as the pound’s decline from ~90% to ~11% of reserves showed, reserve status is not permanent and can erode for years before breaking suddenly.

Debasement vs. deflation tension: The AI transition pushes toward inflation (governments printing to fund crisis) and deflation (AI driving costs down) simultaneously. The net effect is uncertain — and that very unpredictability is itself a risk to anyone whose security is denominated in money.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio, who writes from a free-market, sound-money perspective and aims for independent analysis over political allegiance. Nothing here is financial advice; it is an attempt to think clearly about what money actually is.

This is the second essay in a sub-series on the financial system. Next: who absorbs the loss when debts go bad — student loans, mortgages, and the question of what happens when the promises can’t all be kept.

Subscribe to The Human Transition on Substack to follow the whole series.

Sources and further reading:

• IMF COFER data — dollar share of global reserves below 57% (2025–2026), down from ~72% in 2001

• J.P. Morgan — “De-dollarization: The end of dollar dominance?” (gold accumulation, reserve trends)

• World Gold Council — record central bank gold purchases, 2025–2026

• Federal Reserve — “The International Role of the U.S. Dollar,” 2025 edition

• Historical reference: sterling’s decline from ~90% to ~11% of global reserves, 1947–1973

• US News, FXStreet — de-dollarization analysis and central bank gold buying, 2026

This essay begins a series about the financial system, and I want to open it with the idea that the rest of the series depends on — an idea that runs against almost everything we are taught about money and security. It is this: there is no such thing as an asset that is safe in itself. Every asset you can own is a claim on a functioning system, and it is worth something only as long as that system keeps working. When we forget this — and we are trained to forget it — we mistake a promise for a possession.

Before I make the argument, a word about where I am coming from, because in a series about the financial system you are entitled to know. By instinct and conviction I favor free markets, capitalism, and limited government. I am not a critic of wealth or of those who build it, and nothing here is an argument that owning assets is foolish or that people should not try to secure their futures. The opposite, in fact. This series exists because I take security seriously — seriously enough to want to look hard at whether the things we have been told are safe actually are, in the specific and unprecedented situation the AI transition is creating. What follows is not ideology. It is an attempt to look at the machinery plainly and trace where it leads, including to conclusions I would rather not reach.

And one more honest admission up front, because it shapes everything: it is not at all obvious to me that any political program — mine or anyone else’s — can simply solve what is coming. That uncertainty is not a dodge. It is the most truthful starting point I can offer.

The Thing We Forget About What We Own

Here is the mental trick that money plays on us. We think of an asset as a thing we have — solid, in our possession, ours. A house. A portfolio. A bond. A piece of land. The word “asset” itself suggests something concrete, something you could put your hand on.

But almost nothing about an asset’s value lives in the asset itself. Its value lives in a relationship — a claim on something outside it, honored by a system that has to keep functioning for the claim to hold. Strip the functioning system away, and the thing in your hand is just that: a thing, often a surprisingly useless one. We do not see this in normal times because the systems are working, quietly and invisibly, holding up the value of everything. We only see it when they fail. And the entire argument of this series is that the AI transition is a force capable of straining those systems in ways we have not seriously reckoned with — which means it is worth seeing clearly now, while the systems still hold, rather than discovering it the hard way.

Let me walk through what we own, one kind at a time, and show you the promise hiding inside each possession.

The Rental Property With No One to Rent To

Start with something that feels as solid as ownership gets: real estate. You own an apartment building, or a second home you let out, or a commercial unit. This is wealth you can stand inside. It feels like the opposite of an abstraction.

But what is it actually worth? Its value rests almost entirely on one assumption: that there exists a population of people with enough income to pay you rent. A rental property is not fundamentally a building. It is a claim on the future earnings of tenants. The building is just the mechanism through which that claim is collected.

Now run the thesis of this entire platform through that fact. If the AI transition does what the evidence suggests it may — erode the incomes of broad swaths of ordinary people, the renters, the wage-earners, the very population a landlord depends on — then who, exactly, rents the apartments? The building still stands. The deed is still in your name. But its worth was never in the bricks; it was in the solvency of the people meant to pay you, and that solvency is precisely what the transition threatens. A landlord in a region where incomes have collapsed does not own a gold mine. He owns a maintenance liability with a tax bill. The asset did not change. The system underneath it did.

The same logic, incidentally, is already visible at the top end of commercial real estate, where empty office towers have shown how fast a “premium asset” becomes a stranded one when the human activity it depended on simply stops showing up. Value evaporated not because the buildings changed, but because the assumption beneath them did.

The Stocks With No Customers Left

Now consider equities — shares in companies, the engine of most retirement savings and most serious wealth. Surely this is real ownership: you own a piece of a productive enterprise.

You do. But what makes that piece valuable? A share is a claim on a company’s future profits, and a company’s profits are a claim on the spending of customers who have money to spend. Follow the chain down and every stock portfolio rests, ultimately, on the same foundation as the rental property: a population of people with enough income to buy things. Consumer spending is roughly two-thirds of the economy. Corporate earnings are downstream of household incomes. The stock market is, at bottom, a bet that ordinary people will keep being able to afford things.

So here is the uncomfortable question the transition forces. What happens to corporate earnings — and to the share prices built on them — if AI hollows out the incomes of the consumer base those earnings depend on? A company can cut its own workers and book a short-term gain. But it cannot prosper for long selling to a population that no longer has wages. One firm’s labor cost is another firm’s customer. If enough companies cut at once, they collectively saw off the branch the whole market sits on.

And here the present moment adds a genuinely unsettling wrinkle, one I want to state carefully because it cuts in a strange direction. The market right now is extraordinarily concentrated in the handful of companies building AI itself. The top ten US companies now make up around 40% of the entire S&P 500; the largest few command a value larger than every other stock market on Earth combined; the degree of concentration matches the peak of the dot-com bubble, and total US stock market value sits near twice the size of the whole US economy. What this means is subtle and important: the stock market and the real economy are quietly decoupling. The market can keep rising on the promise of AI even as the broad economy that supports ordinary livelihoods weakens beneath it — for a while. The index looks healthy because a few enormous AI-driven companies are carrying it, which can mask the erosion underneath until the erosion reaches the AI companies’ own customers. A market this concentrated is not the diversified safety people imagine they own. It is a concentrated bet on the very force disrupting everyone else — and when a concentrated bet unwinds, it unwinds fast. The last time concentration looked like this, the unwinding erased some six trillion dollars, around 60% of GDP, and took seven years to recover. I am not predicting that. I am pointing out that “I own a broad stock portfolio” means something very different, and far less diversified, than it did a decade ago.

The Bonds From a Government Under Strain

So perhaps you retreat to the asset traditionally called “risk-free”: government bonds. You lend money to the state, and the state promises to pay you back. For generations this has been the bedrock beneath every cautious portfolio — the safe harbor.

But look at what the promise actually depends on. A government bond is only as good as two things: the government’s ability to pay, and the value of the currency it pays you in. Both are quietly conditional, and the transition presses on both.

Consider the ability to pay. Government solvency rests on the tax base — and the tax base rests, overwhelmingly, on the incomes and consumption of working people. If the AI transition erodes broad employment income, it erodes the very revenue stream governments depend on, at the same moment it dramatically increases the demands onthem for support. Global government debt has already surged past one hundred trillion dollars. A government managing a genuine displacement crisis — falling revenues, soaring obligations — is not an obviously reliable counterparty. It may inflate its way out, or restructure, or be forced into choices that quietly impair the value of the bonds it issued in calmer times.

And consider the currency, which is the deeper issue and the subject of the next essay in this series. A bond is a promise to pay you a certain number of units of money. But the value of those units is itself not fixed — it too is a claim on a functioning system, the most fundamental one of all. If confidence in the currency erodes, the bond can be repaid in full and still leave you poorer, because the units you are repaid in buy less. The “risk-free” asset is risk-free only in nominal terms, and only as long as the system behind the money holds. We will take that apart properly in the next piece, because it is the foundation under all theothers.

The Land They Can Take, or the Hungry Can Storm

So you go further still, to the asset that feels most primal and most real: land. Productive land. Farmland. The thing human beings have treated as ultimate security for all of history— you cannot print it, you cannot delete it, and in the worst case you can grow food on it. Of all the assets, surely this one is safe in itself.

It is the strongest of them, and I will not pretend otherwise. If I were ranking durability, productive land sits near the top. But even here — even here — the safety is conditional, and it is worth seeing why, because it reveals the deepest truth of the whole argument.

Your ownership of land is not a physical fact. It is a legal and social agreement, enforced by the state and respected by your neighbors. Take away either of those, and your “ownership” becomes a piece of paper and a hope. Consider the two ways even land fails. The first is the state itself: governments retain, and have always exercised in moments of crisis, the power to seize, requisition, or repurpose private land — for “public safety,” for “national security,” for the emergency of the moment. A government managing a societal collapse is exactly the kind of government that reaches for such powers. The second is more elemental: in a genuine breakdown, where enough people are desperate and hungry, the productive land surrounded by need does not become a fortress. It becomes a target. Your title means nothing to a crowd that has nothing to lose, and the state that would normally enforce it is the same state that has stopped functioning. This is the same hard truth I have written about before in the context of food — the lone producer of something valuable, in a collapsed world, is not safe but exposed. Land is the most resilient asset precisely up to the point where the social order holds, and not one inch past it.

The Pattern, and What It Actually Means

Step back and the pattern is undeniable. Rental property depends on solvent tenants. Stocks depend on solvent customers. Bonds depend on a solvent government and a sound currency. Land depends on an enforced legal order and a peaceful enough society. Every single asset, no matter how solid it feels in the hand, is at bottom a claim on a functioning system — and the AI transition is, more than anything else, a force capable of straining the systems all of them rest on at once.

I want to be careful and precise about what this does and does not mean, because the lazyreadings in both directions are wrong.

It does not mean you should sell everything, or that owning assets is pointless, or that some collapse is foretold. I am not predicting catastrophe; I am dismantling the illusion of unconditional safety. Assets remain worth owning. Diversification still matters — arguably more than ever. The point is not “own nothing.” The point is to see clearly that what wecasually call “safety” is in fact a bet on continuity — on the systems staying up — and that the AI transition is precisely the kind of force that makes the continuity of those systems less certain than it has been in living memory. The wealthy are not exempt from this. They are simply holding larger and more elaborate versions of the same conditional promises, and are often the most confident that those promises are unbreakable — which is exactly the confidence that the falsely-secure cannot afford.

And here the honesty I promised at the start has to be paid out, because the natural next question is: so what should be done about it? And the truthful answer is that no political tradition has an obvious solution. The free-market reflex — let prices adjust, let capital reallocate, the system will clear — assumes the disruption is the normal kind that markets have always absorbed, when the whole premise of this series is that it may not be. The interventionist reflex — have the state manage it, redistribute, prop up the systems — runs straight into the questions of whether the state is solvent enough, fast enough, or trustworthy enough with that much power to do it without creating new catastrophes. I hold my own market-leaning instincts, and I will tell you plainly: I do not find in them a confident answer to this. That is not a comfortable thing to admit. It is, I think, a more honest place to start than the false certainty being sold from every direction.

What this leaves us with is not despair but reorientation. If no individual asset is unconditionally safe, then real security was never going to be found purely in what you own. It is found, to the degree it can be found at all, in the things this whole platform keeps returning to: the resilience of the broader system, which is worth fighting to preserve; and the strength of the community you are embedded in, which is the one form of security that does not evaporate when the formal systems strain — because it is not a claim on a system at all. It is a claim on each other. That is the asset no transition can repossess, and it is the one almost no one is told to build.

We will spend the coming essays taking the financial system apart piece by piece —currencies, debt, housing, banks, and the things people flee to when they sense all this, like gold and crypto. The aim is never to frighten you. It is to replace the false comfort of assumed safety with the real, usable clarity of knowing what your security actually rests on — so you can build it where it can actually be built.

The train is moving. And the people most certain it cannot reach them are often the ones holding the most elaborate promises about why.

Key Terms Explained

Asset as a claim: The central idea of this piece — that an asset’s value does not reside in the asset itself but in a claim on a functioning system (tenants’ income, customers’ spending, a government’s solvency, an enforced legal order). When the system strains, the value can evaporate while the object remains.

Decoupling of the market from the economy: The phenomenon, acute in 2026, by which a stock market dominated by a few enormous AI companies can rise even as the broad economy supporting ordinary incomes weakens — masking underlying erosion until it reaches the dominant companies’ own customer base.

Risk-free asset (and its conditions): The traditional label for government bonds. This piece shows the label holds only nominally and only while two systems hold: the government’s solvency (its tax base) and the value of its currency.

Conditional safety: The recognition that every asset’s safety is contingent on the continuity of the systems beneath it. Not an argument against owning assets, but against mistaking a bet on continuity for unconditional security.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio, who writes from a free-market, limited-government perspective and aims for independent analysis over political allegiance. Nothing here is financial advice; it is an attempt to think clearly about where value actually comes from.

This is the opening essay of a sub-series on the financial system. Next: the currency itself —the system beneath all the others.

Subscribe to The Human Transition on Substack to follow the whole series.

Sources and further reading:

• Oliver Wyman — “The Impact of an AI Bubble Burst on Global Financial Markets,” 2026 (market concentration, dot-com comparison, equity-to-GDP figures)

• Goldman Sachs Asset Management — 2026 investment outlook (top-10 concentration ~40% of S&P 500; global government debt past $100 trillion)

• BlackRock — “Investing in 2026: AI, War, and Income” (top-10 US stocks’ scale; software-sector repricing)

• Vanguard — 2026 economic and market outlook (valuation percentiles; AI capex concentration)

• US Federal Reserve — Financial Stability Report, May 2026; Governor Barr speech on AI and financial stability, February 2026

Two of the richest men who have ever lived have some advice for you about the AI transition, and it is worth listening to — not because it is reliable, but because of what it reveals.

Elon Musk says you should not bother saving for retirement. “Don’t worry about squirreling money away for retirement in 10 or 20 years,” he said on a podcast earlier this year. “It won’t matter.” In his telling, a “supersonic tsunami” of AI and robotics is about to abolish scarcity itself, ushering in not universal basic income but “universal high income” — a world where, as he puts it, there will be “no poverty” and “no need to save money,” where goods and services arrive at near-zero cost and any job you do is essentially a hobby.

Jeff Bezos has a complementary message: be happy. “If you’ve been digging out a basement for your house with a shovel and somebody’s about to hand you a bulldozer, you should be so happy,” he told an interviewer in May. Far from making people redundant, he insists, AI will “elevate” everyone and even create a labor shortage, because “we have an endless set of things to invent,” and human imagination, freed by machines that handle the execution, will generate more work than there are people to do it.

I want to take these claims seriously, because they deserve a real response rather than a sneer. And I want to be fair: there is a version of each that contains a genuine insight. But I also want to say plainly what almost no one with a platform will say to these men directly: it is remarkably easy to counsel calm about a storm when you are the one selling umbrellas, and standing on the highest ground in the world.

Let me separate what is true in these promises from what is convenient — and then let me tell you why, even if they turn out to be right, you cannot afford to live as though they are.

The Kernel of Truth, Stated Fairly

Start with what is genuinely reasonable in what they say, because this platform does not deal in caricature.

Bezos is drawing on real economic history, and it is not nothing. Across previous waves of technology — the agricultural revolution, the industrial revolution, the computing revolution — the predictions of permanent mass unemployment repeatedly failed to come true. Each time, the technology destroyed categories of work and then, often, created new ones no one had imagined, frequently more than it destroyed. Human wants turned out to be effectively bottomless: when tools made existing things cheaper, people did not stop wanting; they wanted new things, and the work migrated. Bezos is betting AI follows that pattern, and a serious person cannot dismiss the possibility, because it has happened before.

Musk’s underlying point also contains something real. If AI and robotics genuinely drove the cost of producing most goods toward zero, then the relationship between money, work, and survival really would change profoundly. In a true post-scarcity world, the logic of saving for a future of constrained resources would indeed weaken. The internal logic is not insane. It describes a possible end-state.

So I am not here to tell you these men are simply fools or liars. The futures they describe are possible. If everything breaks the right way — if the productivity gains are as vast as they hope, if the new work materializes as it has before, if the costs really do collapse — then some version of their optimism could be vindicated, and I would be glad to be wrong.

But notice how much is hiding in that “if.” And notice who is asking you to bet your one life on it.

What the View From the Top Conveniently Leaves Out

Here is where the reasonable kernel gives way to something that should make you wary, because the gap between what these men say and what they leave out is enormous — and the omissions all point in the same direction.

The first omission is the path. Even if Bezos is entirely right that the destination is more jobs and more abundance, he is conspicuously quiet about the journey. Every previous technological transition that “created more jobs in the end” also inflicted decades of brutal dislocation on the people caught in the middle. The handloom weavers who were displaced by the power loom did not get to enjoy the eventual prosperity of the industrial age — they got impoverished, and many of them died before the supposed compensation arrived. “It works out in the long run” is cold comfort to a person who has to pay rent in the short run. Bezos talks about the bulldozer; he does not talk much about the people standing where the basement used to be. And tellingly, as he delivered this message, his own former company has cut tens of thousands of roles while accelerating automation. The bulldozer is real. So are the people it is digging through.

The second omission is who owns the machines. Musk’s “universal high income” floats in midair, untethered from any mechanism. Who produces this infinite abundance? Machines. Who owns the machines? Not you. The entire vision depends on the owners of the automated economy choosing to distribute its output broadly rather than capturing it —and there is no worked-out policy, no political settlement, no enforceable mechanism by which that sharing happens. It is asserted, not designed. And it is asserted by a man who owns an enormous share of exactly those machines. “Don’t worry, there will be abundancefor all” means something very different coming from the person who will own the factories of that abundance than it does coming from the person who will have to hope for a share of it.

The third omission is the timing, and this is the cruelest one. Even in the most optimistic reading, Musk and Bezos are describing a destination that is, on their own timelines, ten, twenty, thirty years out. But your bills are monthly. Your retirement, if you are in your forties or fifties, is not a distant abstraction — it is the thing you are supposed to be building right now. When the richest man alive tells an ordinary person to stop saving for retirement because abundance is coming, he is asking that person to make an irreversible bet, with their actual survival, on his speculative timeline. If he is wrong — or even just early — he loses nothing. He has more money than he could spend in a thousand lifetimes regardless. The ordinary person who took his advice loses everything: the savings not made, the years not compounded, the security not built. The downside is not shared. It never is.

The Asymmetry That Should Stop You Cold

This is the heart of it, and it is the thing to hold onto when a billionaire tells you to relax.

These men are advising you from a position where they cannot lose. If the optimistic future arrives, they win enormously — they own the technology bringing it. If it does not arrive, or arrives slowly and painfully, they still win, because they already have everything. Their wealth insulates them from being wrong. There is no version of the future in which ElonMusk or Jeff Bezos personally suffers for having been too optimistic about AI.

You do not have that insulation. For you, the difference between their optimism being right and being wrong is the difference between a comfortable life and a catastrophic one. You are being asked to absorb all the risk of a prediction made by people who bear none of it. That is not advice between equals. It is the powerful telling the exposed not to worry aboutthe storm, from inside a house the storm cannot reach.

And there is a further, quieter problem with the advice itself. It is demobilizing. If you believe abundance is inevitable and just around the corner, what do you do? Nothing. You don’t save, you don’t organize, you don’t demand that the transition be managed fairly, you don’t push for the policies that might actually distribute the gains — because why would you, if it is all going to work out automatically? The optimism, whether or not it is sincere, has the convenient effect of keeping you passive precisely when your active engagementmatters most. A population that believes the future is guaranteed to be good is a population that will not fight to make it so. The people who own the machines have every reason to prefer that you wait quietly for the abundance rather than organizing to claim a share of it.

What It Actually Looks Like From Here

Step out of the ivory tower and look at the transition from where most people actually stand— the baker, the stonemason, the freelancer, the surgeon, the architect — and it does not look like a bulldozer being kindly handed over. It looks like uncertainty, and it looks different at every level, but it is uncertain at every level.

The baker watches automated production and delivery reshape the economics of the small shop. The stonemason wonders how far robotics and prefabrication will go. The freelancer— and I write this as one — watches the inquiries slow and the rates come under pressure, with no announcement and no statistic to mark it. The surgeon reads about diagnostic and procedural AI and wonders which parts of a career’s hard-won expertise will still be scarce in fifteen years. The architect sees generative design tools doing in minutes what used to take a team a week. None of these people are being “elevated” in any way they can currently feel. They are being made to wonder, quietly, whether the ground beneath their particular profession is as solid as they assumed.

And here is what none of them can responsibly do: stop saving, stop planning, stop building security, on the say-so of a billionaire. Because the honest truth is that we do not know which future will arrive. Maybe Bezos is right and the new work materializes. Maybe Musk is right and scarcity fades. Or maybe the productivity gains flow almost entirely to the owners, the new jobs do not appear fast enough, the costs do not fall the way the slogans promise, and ordinary people are left more exposed than ever — having been told, all the while, not to worry.

In a world of genuine uncertainty, the rational response is not to adopt the confident optimism of the people who profit from your calm. It is to prepare for a range of outcomes — including the ones the billionaires are not pricing in, because they will never personally have to.

Hope Is Not a Plan, Especially Someone Else’s

I want to be careful not to replace their false certainty with a false despair of my own. The optimistic future is genuinely possible, and I hope it comes. If the costs of living truly collapse and the gains are truly shared, that would be one of the great achievements of human history, and I would celebrate it.

But hope is not a plan, and someone else’s hope — particularly when that someone profitsfrom your believing it — is the least reliable plan of all. The appropriate response to Musk and Bezos is not to call them villains. It is to notice, calmly, that their advice costs them nothing and could cost you everything, and to weight it accordingly.

So save anyway. Plan anyway. Build your security and your community anyway. Demand, anyway, that whoever ends up owning the machines is made to share their output — because that sharing will not happen by itself, no matter how confidently it is promised. If the abundance arrives, you will have lost nothing by being prepared. If it does not, your preparation will be the only thing standing between you and the consequences of having believed a comfortable man on a stage.

The people telling you not to worry will be fine either way. That is precisely why you should not take your cue from them.

The train is moving. And the men who own the train are leaning out the window, telling everyone still standing on the track that there’s nothing to worry about.

Key Terms Explained

Universal high income (vs universal basic income): Musk’s term for a future in which AI-driven abundance provides not just a survival-level safety net but effectively limitless access to goods and services. Distinguished from UBI by its assumption of near-total post-scarcity — an assumption with no current mechanism behind it.

The abundance argument: The claim, advanced by Bezos and others, that AI will drive the cost of goods and services toward zero and generate more new work than it destroys, as previous technological revolutions eventually did. Historically grounded but silent on the dislocation of the transition and on who captures the gains.

The risk asymmetry: The core problem with billionaire optimism — those advising calm are insulated from being wrong, while those taking the advice bear the full downside. The advice costs the adviser nothing and could cost the recipient everything.

Demobilization: The effect by which a belief in inevitable, automatic abundance discourages the saving, organizing, and political demand that would actually be needed to secure a fair outcome — conveniently keeping the exposed passive while the powerful proceed.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio. This piece critiques the public arguments of named individuals on their merits; it is not a personal attack, and it takes no partisan side.

Subscribe to The Human Transition on Substack to follow the whole series.

Sources and further reading:

• Fortune — “Elon Musk says saving for retirement is irrelevant because AI will create a world of abundance,” January 2026

• Fox Business — “Musk predicts AI will create ‘universal high income’ and make saving money unnecessary,” December 2025

• The Daily Economy — “AI Abundance Won’t End Inflation, Nor Make Money Meaningless,” April 2026 (critique of the monetary argument)

• Fortune — “’AI is going to create a labor shortage’: Jeff Bezos,” June 2026 (VivaTech remarks)

• Gizmodo / Entrepreneur — Bezos “bulldozer” and “be so happy” remarks, CNBC interview, May 2026

• The Hill / TheStreet — Bezos labor-shortage claims alongside Amazon’s automation-driven job cuts, 2026

• Challenger, Gray & Christmas; Goldman Sachs — 2026 layoff and AI-displacement data (background)

A few days ago, something happened that this platform has been circling for its entire existence. OpenAI proposed handing the United States government a five percent equity stake in the company — worth roughly forty-two billion dollars at its current valuation — as the seed of what it calls a “Public Wealth Fund,” a vehicle modeled on things like the Alaska Permanent Fund that would, in theory, distribute some of the wealth AI generates back to ordinary citizens. The proposal envisions other major AI labs contributing similar stakes. It has been discussed at the highest levels of government. And it is being framed, by its champion, as the best way to let the public share in the upside of artificial intelligence.

This is genuinely significant, and I want to treat it with the seriousness it deserves — which means neither cheering it as generosity nor dismissing it as a scam, but doing the harder thing: seeing clearly what it is, what it could become, and why the way it is being handled tells you something important about the whole transition. Because underneath the headline number lies the single most important question this entire platform keeps arriving at, and it is finally being asked out loud, at scale, in a room where decisions actually get made.

Let me start with why this matters so much, and then get to what troubles me about it.

The Right Question, Finally Being Asked

For a long time, the conversation about AI and ordinary people has been stuck in diagnosis — this job will go, that industry will change, this is coming. The question that logically follows, the one that actually matters for how people will live, has mostly been left hanging: if AI generates enormous wealth while displacing the labor that used to distribute that wealth through wages, how does any of it reach the people whose livelihoods it disrupts?

I wrote, recently, about a small nonprofit experiment — a program sending a thousand dollars a month to a few dozen workers displaced by AI, funded by donations. It was a decent, humane effort, and I used it to draw a distinction I think matters more than almost any other in this whole subject: the difference between charity and a claim. Charity is a voluntary gift, given at the giver’s discretion, revocable, and accompanied by gratitude owed in return. A claim is a structural, enforceable right — something owed to you, that does not depend on anyone’s goodwill. The whole future of how ordinary people fare in this transition, I argued, turns on which of those two models we build.

What is striking about OpenAI’s proposal is that it takes that exact question — how does AI wealth reach the public — and escalates it from a three-hundred-thousand-dollar charity pilot to a forty-two-billion-dollar proposal negotiated with the President, the Treasury, and the Commerce Department. The scale has changed entirely. The question has gone from the margins to the center. And the mechanism being floated — a sovereign-wealth-fund model that gives the public an ownership stake and distributes the returns — is, in its structure, much closer to the claim side of my distinction than the charity side. A citizen who owns a share of the wealth, and receives dividends from it as a right, is in a fundamentally stronger position than one who receives a discretionary handout. So on its face, this is movement in a genuinely important and potentially good direction. I want to be honest about that before I complicate it. If it were real, enforceable, and well-designed, a public stake in AI’s wealth could be one of the more serious answers anyone has put forward to the central economic problem of this transition.

That “if” is doing an enormous amount of work. Let me get to it.

The Tell: It’s a Donation, Offered to Manage Blowback

Here is the first thing to notice, and it is right there in the reporting, stated plainly: the proposal is described as a donation, and its purpose, according to the people describing it, is partly to “secure good relations with the administration” and to “address political blowback.” Sit with that framing, because it changes the character of what is being offered.

This is not the public claiming a right that is owed to it. It is a company offering a gift — voluntarily, on its own terms, at a moment when it faces mounting political pressure, souring public sentiment, regulatory threats, and the recent experience of watching the government reach into the industry and switch off a competitor’s models. In that context, a voluntary offer of equity is not only, and maybe not even primarily, an act of sharing. It is an act of managing — managing the relationship with the state, managing the political risk, managing the narrative around who benefits from AI at exactly the moment that narrative is turning hostile. A gift offered to defuse blowback is a very different thing from a right the public can enforce, even if the dollar figure is identical.

And because it is a gift rather than a right, it carries all the properties of a gift that I warned about in the context of that earlier charity program. It is given at the giver’s discretion, which means it can be designed, on the giver’s terms. It can be structured to vest only at some future milestone — an IPO, a revenue threshold — rather than now, which means it may be more promise than transfer. Its governance can be arranged so the public holds the economic upside but none of the control. And crucially, by offering it, the company gets to stand as the generous author of the solution to a problem it is helping to create — buying, with a stake it designed, the moral high ground and the political goodwill that a mandatory claim would never have conferred. The very voluntariness that makes it look generous is also what makes it powerful for the giver, because a gift earns gratitude and buys narrative in a way that paying what you owe never does.

I am not saying the proposal is therefore worthless or purely cynical. I am saying that the framing — donation, offered to manage political blowback — is exactly the framing my earlier analysis predicted would appear, and it is the framing you should examine most carefully, precisely because it is the one designed to disarm your scrutiny by looking like generosity.

The Empty Chair, Again — Now With the Public’s Own Share

Here is the part that I find most revealing, and it connects this moment to a pattern I have traced across several pieces on this platform.

Look at who is in the room where this is being decided. OpenAI. The White House. The Treasury Secretary. The Commerce Secretary. Rival lawmakers with competing proposals of their own. The heads of the other AI labs who might be asked to contribute. Every party to this negotiation is either a powerful company or a powerful arm of the state. And the beneficiary — “the American public,” the ordinary citizens whose share is being discussed — is not at the table. They are the subject of the conversation, not a participant in it. Their share is being designed for them, by the company offering it and the government receiving it, according to terms they have no direct voice in setting.

This is the same empty chair I have written about before — in the story of the government switching off Anthropic’s models, a fight between a company and a state with the public as spectator; in the story of one AI CEO attacking his rivals, a fight among giants with the public as the territory fought over. And here it is again, in perhaps its purest form yet: even the generosity, even the mechanism explicitly meant to benefit ordinary people, is being designed over their heads. The public’s own share of AI’s wealth is being negotiated in a room the public is not in. That should tell you something structural about this entire era: the ordinary person is so absent from the decisions that shape their future that they are absent even from the decisions about their own benefit. The gift is being wrapped for them by people they did not choose, on terms they cannot see.

And this matters practically, not just symbolically, because the terms are everything. Whether this becomes a genuine, enforceable, adequate, well-governed public claim on AI wealth — or a token stake, vesting on distant milestones, controlled by others, distributing trivial sums while buying enormous political cover — depends entirely on details being decided right now, by parties whose interests do not perfectly align with the public’s. Whether the fund pays a meaningful dividend or a symbolic one; whether the stake is real ownership or a passive holding with no control; whether it vests now or only if the company chooses to go public; whether other labs genuinely match it or quietly decline — these are the questions that determine whether this is a historic step or a sophisticated piece of public relations. And they are being answered without the beneficiaries in the room.

The Honest Part, and What to Take From It

Let me hold the honest balance, because this is exactly the kind of moment where it would be easy to collapse into one of two lazy positions, and both would be wrong.

The cheerleader says: a company is voluntarily offering forty-two billion dollars to the public — this is wonderful, be grateful, the system is working. The cynic says: it is a transparent bribe to buy political protection, meaning nothing, worth less. And the honest position refuses both, because both are partly true and neither is the whole. It genuinely could be a meaningful step toward sharing AI’s upside — the sovereign-wealth-fund model is a serious idea with real precedent, and a public that owns a piece of the AI economy is better off than one that owns nothing. And it is also, simultaneously, a voluntary, giver-designed, blowback-managing gift whose generous framing is itself a form of power, offered on terms the public had no part in setting. Both of those are true at once, and the skill this era demands is holding them together rather than grabbing the one that fits your mood.

Where do I land, with my own instincts disclosed? My free-market lean makes me genuinely wary of the government taking equity stakes in private companies — it is a strange and potentially troubling arrangement, entangling the state and the firms it is supposed to regulate in ways that could corrupt both. And my skepticism of concentrated power makes me uneasy about any mechanism where the powerful get to design the terms of their own accountability. So I am not here to endorse this proposal. But I am also not going to pretend the underlying question is illegitimate, because it is the most legitimate question there is: how does the wealth of this transition reach the people it displaces? OpenAI has, whatever its motives, put a real answer on the table, and the answer deserves neither reflexive applause nor reflexive dismissal — it deserves scrutiny, on exactly the terms that determine whether it is a claim or a costume.

For you, the usable insight is this. When you hear that AI companies are proposing to share their wealth with the public, do not stop at the headline number, because the number is the easy part and the terms are everything. Ask the questions the beneficiaries would ask if they were in the room: Is it a right I can enforce, or a gift I must be grateful for? Does it vest now, or on some distant promise? Do I own it, or merely receive from it at someone’s discretion? Is the sum meaningful, or symbolic? And who designed these terms — me, or the people who benefit from my accepting them? These are the questions that separate a genuine claim from a sophisticated piece of narrative management. And the fact that they are being answered right now, without the public in the room, is exactly why the public — why you — should be paying the closest possible attention to this, of all things. Because this may be the single most important thing the transition ever offers ordinary people, and it is being wrapped, right now, by hands that are not theirs.

The train is moving. And someone has just offered the passengers a share of the ticket revenue — a real and serious offer, worth taking seriously — designed entirely by the people who own the train, for passengers who were never invited to the meeting.

Key Terms Explained

The Public Wealth Fund proposal: OpenAI’s floated offer to place ~5% of its equity (roughly $42.6 billion at its ~$852B valuation) into a government-held, sovereign-wealth-fund-style vehicle modeled on the Alaska Permanent Fund, intended to distribute AI-generated returns to the public. Other labs would be asked to contribute similar stakes. As of writing, a proposal — not a signed deal — and likely to require congressional approval.

Charity vs. claim (revisited): The distinction at the heart of this piece. A voluntary donation is given at the giver’s discretion, revocable, earning gratitude and narrative control. An enforceable claim is a right owed regardless of goodwill. The proposal’s sovereign-fund structure leans toward “claim,” but its framing as a voluntary donation to manage political blowback carries the properties — and dangers — of a gift.

The blowback tell: The reporting frames the offer as partly meant to “secure good relations with the administration” and “address political blowback” — signaling that the gift is also an act of managing political and regulatory risk, and buying, with a self-designed stake, the goodwill and moral standing a mandatory claim never would.

The design questions that decide everything: Whether the stake vests now or on distant milestones; whether it is real ownership or a passive holding; whether it pays a meaningful or symbolic dividend; whether other labs genuinely match it — the terms, set right now without the public in the room, that determine whether this is a historic public claim or sophisticated public relations.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio, who writes from a free-market, limited-government perspective and aims for independent analysis over political allegiance — including unease about government equity in private firms, and about the powerful designing the terms of their own accountability. Nothing here is financial advice.

This is a standalone current-events piece. Subscribe to The Human Transition on Substack to follow the whole series.

Sources and further reading:

• Financial Times — report that OpenAI proposed handing the US government ~5% equity via a “Public Wealth Fund,” July 2, 2026 (donation framed to “secure good relations” and “address political blowback”)

• CNBC — “OpenAI proposes U.S. government own 5% stake to address political blowback,” July 2, 2026 (~$42.6B at $852B valuation; precedents: Intel 10% stake, Nvidia/AMD China chip revenue shares)

• TechCrunch — “OpenAI proposed donating 5% of its equity to a US sovereign wealth fund,” July 2, 2026 (other labs to contribute; likely congressional approval required; talks preliminary)

• OpenAI — “Industrial Policy for the Intelligence Age” white paper, April 2026 (original public-wealth-fund concept; returns “distributed directly to citizens”)

• Reporting on structure and open questions (governance, vesting, distribution formula, antitrust) and on Sen. Sanders’ competing equity-tax proposal, July 2026

• Alaska Permanent Fund — model precedent (~$91B in assets; annual resident dividends since 1982)

Something strange is happening in the numbers, and it is worth slowing down to look at it directly, because both the people shouting that AI is ending all work and the people insisting it is all hype are missing what the evidence actually shows.

Here is the strange thing. Across 2026, companies have been shedding workers at a punishing pace — on the order of a thousand a day in tech, finance, and healthcare, nearly double last year’s rate. And more than half of those layoff announcements name the same culprit: artificial intelligence. On the surface, this looks like the great displacement arriving exactly as predicted. The robots came, the jobs went, the story writes itself.

Except the story does not hold together when you examine it. Because at the very same time, a growing pile of evidence shows that these AI-blamed layoffs are not actually delivering the results AI was supposed to deliver. The companies cutting the most are not performing any better than the companies cutting the least. In several cases, the ones cutting less are doing better. Which raises a question that should make everyone — the optimists and the doomers alike — sit up: if the AI is not actually doing the work, then why, exactly, are these people losing their jobs?

The answer is more troubling than either simple story, and it tells you something important about the world you are navigating.

The Term for It Is “AI Washing”

A word has emerged for the gap between what companies say is causing their layoffs and what is actually causing them: AI washing. It is built on the model of “greenwashing” — the practice of dressing up an ordinary corporate decision in fashionable language to make it more palatable. In this case, the fashionable language is artificial intelligence, and the ordinary decision is cutting payroll.

This is not a fringe accusation from disgruntled workers. It has moved squarely into the mainstream, acknowledged by some of the very people building the technology. The CEO of OpenAI, Sam Altman, said it plainly earlier this year: there is some genuine displacement of specific jobs by AI, and there is also some “AI washing,” where companies blame AI for layoffs they would have carried out anyway. Analysts at Deutsche Bank went further, predicting that AI-redundancy whitewashing would be “a defining feature of 2026” — large companies using AI as a rhetorical shield for cuts actually driven by tariffs, economic uncertainty, and ordinary cost pressure. A Wharton professor and researchers at Forrester have used the same term for the same reason.

Notice what is being admitted here. The people closest to the technology are telling you that a meaningful share of the layoffs attributed to AI are not really about AI at all. They are about something older and more familiar: companies cutting costs and reaching for the most flattering possible explanation.

Why “AI” Is the Perfect Excuse

To understand why a company would blame AI for a layoff it was going to do regardless, you have to understand what the explanation accomplishes. It is, from a corporate communications standpoint, close to perfect.

Think about the alternatives. A company that announces layoffs because it overhired during the boom years is admitting a planning failure. A company that cuts staff because demand is weakening is signalling trouble to its investors. A company that slashes payroll purely to juice quarterly margins looks greedy and short-term. None of these stories flatter anyone.

Now consider the AI story. A company that announces layoffs “because of AI” is telling the market something investors love to hear: we are technologically ahead of the curve, and we are becoming leaner at the same time. Both messages push the stock price in the same direction. Being disrupted by your own foresight sounds like strength. Being forced into cuts by weak demand sounds like weakness. The same act — firing people — can be framed as either, and AI lets you choose the flattering frame.

This is why the explanation has spread so fast and so far. It is not necessarily that AI is doing all these jobs. It is that “AI” has become the socially acceptable, even prestigious, cover story for decisions companies wanted to make anyway. When an explanation is this convenient, you should always look twice.

The Tell: The Layoffs Aren’t Working

Here is the piece of evidence that cuts through the fog, and it comes from a large, careful study by the research firm Gartner — not an advocacy group, but a corporate advisory firm whose clients are the very executives making these decisions.

Gartner surveyed 350 global executives at billion-dollar companies, all of them already piloting or deploying AI. The finding was blunt: about 80 percent had cut headcount, some by as much as 20 percent — and the companies that cut the most showed nearly identical financial returns to the companies that cut the least. In several cases, the ones that cut less actually performed better. The Gartner analyst who led the research put it about as plainly as a corporate advisory firm ever will: chasing value purely through headcount reduction “is likely to lead most organizations down a path of limited returns.”

Sit with what that means. If AI were genuinely replacing these workers — doing their jobs faster and cheaper — the companies cutting hardest would be pulling ahead. They are not.The returns simply are not there. And the pattern is visible in specific, almost comic cases: companies that cut staff and then had to scramble to hire them back. One major firm eliminated hundreds of customer-service roles, watched the quality of its service collapse, and started rehiring. Another automated large parts of its HR operation and reversed course when the systems could not handle the judgment calls real situations require.

So if the layoffs are not paying off, and the AI cannot yet reliably do the work, what is actually going on?

The Darker Reading: Fired on a Bet

Here is the part that I think matters most, and that almost no one is saying clearly.

The most honest interpretation of the evidence is not that AI is replacing these workers today. It is that companies are cutting workers now, based on a bet that AI will be able toreplace them later. The jobs are not being handed to AI systems that are already doing the work. They are being eliminated in anticipation of AI systems that companies believe will eventually do the work. The layoff comes first. The capability is supposed to arrive afterward.

This reframes everything, and it should disturb you more than the simple “robots took the jobs” story, not less. Because it means real people, with real bills and real families, are being let go today not because a machine can do their job, but because their employer is wagering that a machine will be able to someday — and would rather not be paying salaries in the meantime. The worker becomes the thing you cut to free up cash for the technology that is supposed to make the worker unnecessary. You are, in effect, asked to fund your own replacement, in advance, whether or not the replacement ever actually works.

And the cash is going somewhere specific. This is the detail that ties the whole thing together. While these companies shed workers, they are pouring staggering sums into AI infrastructure — the four largest are spending on the order of seven hundred billion dollars in a single year, much of it on AI data centres and chips. The layoffs and the capital spending are two ends of one decision. Payroll is being converted into compute. The people are being traded for the bet.

I want to be careful and fair here, because this platform does not deal in villains. Some of this displacement is real — there are specific roles, especially routine and entry-level ones, where AI genuinely is absorbing the work, and I have written elsewhere about how seriously that should be taken. And some companies are making these decisions in good faith, genuinely restructuring around a technology they believe in. The point is not that every executive is lying. The point is that “AI did it” has become a story doing enormous work to obscure a more ordinary and more uncomfortable reality: workers are bearing the immediate cost of a transition whose benefits, if they arrive at all, will accrue to someone else, later.

Why It Doesn’t Matter Whether It’s “Really” AI — and Why It Matters Enormously

Here is the conclusion that I think cuts through the entire noisy debate, and it has two halves that only look contradictory.

To the worker, it does not matter whether AI really took the job or merely served as the cover story. Either way, the job is gone, the income has stopped, the mortgage is still due. The person laid off “because of AI” and the person laid off because their employer wanted leaner margins and reached for the AI excuse are standing in exactly the same unemployment line. The distinction that obsesses the commentators — was it real displacement or just AI washing? — is invisible from where the displaced person is standing. This is why the endless argument about whether AI is “really” causing job losses misses the human reality entirely. The cause on the press release does not change thesilence in the inbox.

And yet the distinction matters enormously — not for whether you are affected, but for understanding who is deciding your fate and why. Because what AI washing reveals is that your livelihood is, to a degree that should unsettle you, subject to a story. A narrative. A framing chosen in a boardroom and sold to investors. Your job can be eliminated not because a machine can do it, but because eliminating it and saying the magic word “AI” makes a stock go up. That is a kind of power over your life that has nothing to do with the technology’s actual capabilities and everything to do with who gets to decide what counts as a good reason to let you go — and who benefits when they do.

That is the real lesson, and it connects to everything this platform has been documenting. The AI transition is not only a story about what machines can and cannot do. It is a story about power — about who gets to make decisions that reshape millions of lives, what explanations they are permitted to hide behind, and who is left to absorb the consequences while the gains flow upward. Whether the robot can do your job is, in a sense, a side issue. The deeper issue is that the people deciding your future have found a powerful new word to justify whatever they were going to do anyway, and the burden of proof has somehow landed on you to show you are still needed, rather than on them to show that you are not.

What to Take From This

Do not let this make you cynical, and do not let it make you complacent either. Both would be mistakes.

Not complacent, because the displacement underneath the hype is partly real, and the direction is real even where this year’s specific numbers are inflated by convenient storytelling. The pressure on jobs is genuine and it is not going away. If anything, the “fired on a bet” dynamic means you can be displaced before the technology can even do your job— which is a wider exposure, not a narrower one.

But not cynical either, because seeing the AI-washing clearly is itself a kind of power. It means you can read the news about AI and layoffs with clearer eyes — neither swallowing the panic that says the robots have already won, nor the comfort that says it is all just amarket correction. It means you understand that when a company blames AI for cuts, it may be telling the truth, or it may be reaching for the most flattering available excuse, and that the difference tells you something about that company and about the game being played.

And it means you can ask the better questions — the ones this platform keeps coming back to. Not just “can AI do my job?” but “who is deciding, on what basis, and who profits from the answer?” The first question keeps you staring at the technology. The second one shows you the actual machinery of the transition: the boardrooms, the investor calls, the seven-hundred-billion-dollar bets, and the quiet conversion of human livelihoods into the fuel for it.

The robots are not the whole story. They may not even be the main story. The main story is who gets to decide, what they are allowed to say about it, and who pays. Keep your eyes there.

The train is moving. And sometimes the people driving it are blaming the train for choices they made with their own hands on the wheel.

Key Terms Explained

AI washing: Attributing layoffs or business decisions to artificial intelligence when the real drivers are more conventional — overhiring corrections, weak demand, margin targets, or capital reallocation. Named after “greenwashing.” Acknowledged in 2026 by figures including OpenAI’s Sam Altman and analysts at Deutsche Bank and Forrester.

Headcount reduction without ROI: The Gartner finding that companies cutting the most staff under the AI banner showed no better financial returns than those cutting the least — evidence that the layoffs are often not driven by genuine AI productivity gains.

Fired on a bet: The dynamic in which workers are cut now in anticipation of AI capabilities expected later, rather than because AI is already doing their jobs — meaning the layoff precedes the actual replacement, if the replacement ever materialises.

Payroll-to-compute conversion: The pattern of companies reducing labour costs while simultaneously committing enormous sums to AI infrastructure — in effect, trading workers for the bet on the technology meant to replace them.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio. This piece analyses a developing economic story and takes no political side; it reasons toward the structural point and leaves the partisan conclusions to the reader.

Subscribe to The Human Transition on Substack to follow the whole series.

Sources and further reading:

• Gartner — “Autonomous Business and AI Layoffs May Create Budget Room, but Do Not Deliver Returns,” May 2026, and analyst commentary by Helen Poitevin

• Fortune — “AI isn’t paying off in the way companies think,” May 11, 2026

• Yahoo Finance — “Companies Are Firing Workers To Fund AI That Isn’t Working Yet,” June 2026

• Sam Altman, remarks on AI washing, India AI Impact Summit, February 2026

• Deutsche Bank — analyst note on AI redundancy whitewashing, 2026

• Challenger, Gray & Christmas — AI-attributed layoff tracking, 2026

• TechTimes / TrueUp — 2026 layoff pace and AI infrastructure capital expenditure data

• Stanford HAI — 2026 AI Index, entry-level software employment data (background)

Subscribe now

Something genuinely good is happening, and I want to start by honoring it, because what I am going to say next should not be mistaken for cynicism about the people involved.

Earlier this year, a coalition of nonprofits began sending real money to real people hurt by the AI transition. The program is called the AI Dividend. It pays $1,000 a month, for a year, no strings attached, to a small cohort of 25 to 50 workers who have lost pay, jobs, oropportunities to artificial intelligence. It pairs the cash with career support. It was designed with input from hundreds of affected workers. It is run by people who saw the damage upclose — one of the organizers is a veteran software engineer who watched her own programming students graduate into a market that no longer had room for them, landing only “demeaning” work reviewing the output of the machines that had taken the real jobs.

This is decent, humane, useful work. The people doing it are not the problem. For the handful of workers receiving that $1,000 a month, it is not an abstraction — it is rent, food, breathing room, dignity during a frightening time. I would never tell those workers, or those organizers, that what they are doing does not matter. It matters enormously to the people it touches.

And yet. Buried in the design of this well-meaning program is a structural choice that should give us all pause — not because it is malicious, but because if it becomes the model for how society responds to AI displacement, it leads somewhere troubling. The organizers have been candid about how they intend to scale from a few dozen workers to thousands. Their plan is to ask the major AI companies to donate.

Sit with that for a moment, because it contains the whole problem.

The Plan Is to Ask the Disrupters to Be Generous

The AI Dividend currently runs on about $300,000. Its organizers hope to distribute $3 million this year — a tenfold jump — and they have been refreshingly honest that the path to get there runs through the AI companies themselves. They are, in their own words, inactive conversations with AI labs about long-term funding. They have approached firms including Anthropic. And their pitch leans on a hopeful premise: the major AI CEOs, they note, say they believe in supporting displaced workers.

I want to be fair to this instinct, because the underlying logic is not wrong. The wealth that AI is generating is enormous, and the people being displaced have a genuine moral claim on some of it. Routing money from the companies profiting off the disruption toward the workers absorbing its cost is, at the level of basic fairness, exactly right. The instinct — the AI-generated wealth should reach the people AI displaces — is correct. I share it. This entire platform is built on a version of it.

The problem is not the instinct. The problem is the wiring. Because there is a world of difference between two ways of getting money from the people who own the machines tothe people displaced by them. One is a claim — a structural, enforceable, non-negotiable obligation: a tax, a levy, a dividend owed by right. The other is charity — a voluntary gift, given at the discretion of the giver, dependent on their goodwill, revocable at any time, and accompanied, always, by gratitude owed in return. The AI Dividend, by necessity, is being built as the second kind. And the second kind is a trap.

Why Charity From the Disrupter Is Different in Kind, Not Just Degree

It would be easy to say “charity is better than nothing, so what’s the objection?” But that misses something important about what charity does to the relationship between the giver and the receiver — and why, in this particular case, it entrenches the exact problem itappears to solve.

Consider what it means for the displaced to depend on the generosity of the companies that displaced them.

First, it makes the remedy depend on the goodwill of the people with the least incentive to provide it, and the most power to withdraw it. A claim you can enforce. A gift you can only hope for. If AI companies fund displacement relief voluntarily, they fund it when it is convenient, in the amounts they choose, for the recipients they approve of, for exactly as long as it serves their interests — and they stop the moment it does not. The displaced worker’s security becomes contingent on the quarterly mood of a corporation. That is not security. It is dependence dressed up as solidarity.

Second, it transfers not just money but power — in the wrong direction. When the disrupter becomes the benefactor, the displaced person is placed in the position of supplicant: grateful, beholden, and quiet. You do not bite the hand that feeds you. A population that depends on the charity of AI companies for its survival is a population structurally discouraged from organizing against those companies, regulating them, taxing them, or demanding they change. The charity purchases not just goodwill but acquiescence. It converts a potential adversary into a grateful dependent. This is the oldest function of noblesse oblige: the lord’s generosity to the peasants is also the mechanism that keepsthem peasants, and keeps them loyal.

Third — and this is the part that should genuinely worry you — it lets the companies buy the story. If the major AI firms become the visible funders of displacement relief, they get to stand simultaneously as the cause of the problem and the sponsor of the solution. They get to displace you and then be thanked for catching you. The narrative becomes “look how responsibly the AI industry is handling the transition it created” — which is precisely the narrative that forestalls the harder structural responses. Voluntary charity, in this light, is not the opposite of avoiding accountability. It is one of the most effective forms of avoiding accountability. A modest, well-publicized donation is far cheaper than a tax, and it buys something a tax never could: the moral high ground.

There is a name some scholars give to this dynamic — symbolic violence — the way a system of domination gets to dress itself in the language of generosity, so that the people on the losing end experience their subordination as a kindness, and are grateful for it. I do not use that phrase to be inflammatory. I use it because it names something precise: the deepest trap is not being harmed, but being harmed and then taught to say thank you.

The Tell: Even the Optimists Are Building Charity, Not Claims

Here is what makes this moment so clarifying. The AI Dividend’s organizers are not naive, and they are not cynical. They are smart, committed people doing the best they can with the tools available. And the tool available to them — the only one that exists right now — is to ask nicely.

That tells you something about the state of the world. There is no mechanism by which AI-displaced workers can claim a share of AI-generated wealth as a matter of right. No dividend owed. No levy in place. No structural channel at all. So even the people most committed to helping the displaced are forced to route their efforts through the one channel that exists: the voluntary goodwill of the powerful. Not because they think charity is the ideal solution, but because the better solution — an enforceable structural claim —does not yet exist for them to use.

That absence is the actual story. The fact that the most promising real-world effort to help displaced workers has no choice but to pass the hat to the very companies doing the displacing is not a feature of these organizers’ imagination. It is a measure of how completely we have failed, so far, to build anything better. The charity model is not winning because it is good. It is winning because it is the only thing on the field.

What the Better Version Would Look Like

I do not want to only diagnose, so let me be concrete about the distinction that matters, because it is the whole point.

The difference between charity and a claim is the difference between hoping the wealth reaches you and being owed it. A genuine response to AI displacement would not depend on whether a CEO is feeling generous this quarter. It would be structural: a standing, enforceable public claim on the wealth AI generates, paid out as a right rather than a favor. The mechanisms people debate — a tax on automation, a levy on AI productivity gains, a sovereign wealth fund that gives every citizen a real ownership stake in the automated economy, a genuine dividend owed to the public whose data and society made the technology possible — differ in their details and all carry hard questions this platform has never pretended are easy. But they share the one feature that charity can never have: they do not require the permission of the people who would rather not pay.

That is the line that matters. Not “is money reaching displaced workers?” but “does it reach them by right, or by grace?” Because a right preserves your dignity and your power to keep demanding. Grace costs you both.

And notice: building the structural version is exactly the kind of slow, political, collective work that charity quietly discourages. Why fight for a tax when a donation is already arriving? Why organize when you have been told the companies are handling it? The existence of the charitable patch reduces the pressure to build the structural claim — which is, conveniently, precisely the outcome the companies funding the charity would prefer. The patch is not just inadequate. It is, in a quiet way, an obstacle to the cure.

Take the Help. Refuse the Story.

So what is a person to make of all this? Let me be careful, because the wrong conclusion here would be cruel.

If you are a displaced worker and a program like the AI Dividend can help you — take it. Without hesitation and without guilt. The structural critique I have just made does not mean you should refuse rent money on principle while you wait for a better world to arrive.

That would be madness, and it would be the displaced once again paying the price for everyone else’s principles. Take every bit of help you can get. The critique is not aimed at you.

The critique is aimed at the rest of us, and at the story. Take the help — but refuse the story that comes with it. Refuse the narrative that says the AI industry is generously handling the problem it created, that a few thousand donated dollars settles the account, that gratitude is the appropriate response to being caught after being pushed. Hold both things at once: this charity is genuinely helping real people and it must not be allowed to become the model, because the model we actually need is one where you are owed, not given.

The people running the AI Dividend understand this better than anyone — they are reaching for the goodwill of AI companies not because it is right but because it is all they have. Our job, the rest of us, is to build the thing that would make their hat-passing unnecessary: a world in which the wealth AI creates reaches the people it displaces by right, enforceable and permanent, requiring no one’s permission and no one’s thanks.

The train is moving. And right now, the people it has run over are being handed a thousand dollars and asked to be grateful to the people driving it. And only a few people. Take the thousand dollars. Keep your eyes on the train.

Key Terms Explained

The AI Dividend: A 2026 pilot program by the nonprofits the AI Commons Project and What We Will (via the Fund for a Guaranteed Income), paying $1,000 a month for a year to 25–50 workers displaced by AI, with plans to scale by soliciting donations from AI companies.

Charity vs. claim: The central distinction of this piece. Charity is voluntary, discretionary, revocable, and creates dependence and gratitude. A claim (a tax, levy, or dividend owed by right) is enforceable, non-negotiable, and preserves the recipient’s dignity and political power. Money can reach the displaced either way; only one way keeps them free.

Symbolic violence: A concept (from sociologist Pierre Bourdieu) describing how systems of domination can clothe themselves in the language of generosity, so that those on the losing end experience their subordination as kindness and feel grateful for it. Relevant to disrupter-funded relief.

Structural claim mechanisms: Enforceable alternatives to charity — automation taxes, AI productivity levies, citizen sovereign-wealth funds, public dividends — that route AI-generated wealth to the public by right rather than by the goodwill of the companies. Each carries hard design questions, but all share independence from corporate permission.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio. This piece honors the people running real relief programs while critiquing a structural model; it takes no partisan side and reasons toward the distinction it draws.

Subscribe to The Human Transition on Substack to follow the whole series.

Sources and further reading:

• Blood in the Machine (Brian Merchant) — “The first basic income for workers impacted by AI has begun sending out $1,000 monthly payments,” March 2026

• Basic Income Earth Network (BIEN) — coverage of the AI Dividend and “AI is speeding up the deadline for basic income,” 2026

• TechSpot, Gizmodo, Fast Company, Yahoo Finance — AI Dividend reporting, March–April 2026

• Fund for a Guaranteed Income / AI Commons Project — program details and scaling plans

• Bélisle-Pipon, “AI, universal basic income, and power: symbolic violence in the tech elite’s narrative,” 2025 (academic background on the symbolic-violence critique)

Subscribe now

UPDATE (July 1, 2026): The standoff has resolved. After a 19-day shutdown, the Commerce Department lifted the export controls on June 30; Fable 5 returned to global users on July 1, while Mythos 5 was restored only to approved U.S. organizations through Anthropic’s Project Glasswing. The trigger, it later emerged, was a jailbreak technique discovered by Amazon researchers and flagged to the government by Amazon’s CEO — one giant reporting a rival’s flaw. Critics noted the shutdown had handed valuable time to Chinese open-source developers racing to catch up. I’ve left the essay below as originally written, because its point was never the news; it was the structure the news revealed — and that structure held. The models came back the way they went dark: through a negotiation between a company and a government, with the public a spectator to both. The chair stayed empty the whole time. Read on for why that is the part that outlasts the headline.

On Friday morning, hundreds of millions of people could use a piece of software called Claude Fable 5. By Friday night, it was gone. And the company that built it did not make that call.

If you have never heard of Fable 5, that is rather the point of this essay. This is a story that looks, on its surface, like it has nothing to do with you. Three enormous powers — the government of the United States, a $965-billion AI company, and a frontier technology most people will never knowingly touch — are fighting over export-control law, national-security authorities, and a model name that sounds like a children’s book. It reads like dispatches from a control room you will never enter, in a language you were never taught.

I want to argue that you should pay attention anyway. Not because the technical details matter to your Tuesday, but because of what the whole episode quietly reveals: the most consequential decisions about the technology now reshaping your work, your income, and your future are being made in a room you are not in, by powers that did not ask you, and could not be influenced by you even if you tried. The remoteness you feel reading this is not a sign that it doesn’t concern you. It is the concern.

Let me tell you what happened, and then what it means.

What Happened, Plainly

Earlier this month, Anthropic — one of the handful of companies at the absolute frontier of artificial intelligence — released two of its most powerful models yet. Fable 5, made available to the public with safety guardrails built in. And Mythos 5, a more capable, less restricted version for select partners. Both descend from a model family unusually good at finding hidden flaws in software — the kind of capability that is enormously valuable for defending computer systems, and enormously dangerous if pointed the other way.

On Friday, June 12, at 5:21 in the evening, Anthropic received a letter. The directive, issued under export-control authority by the Commerce Secretary, ordered the company to suspend all access to both models for any foreign national — anywhere in the world, inside or outside the United States, including Anthropic’s own foreign employees. Because there is no reliable way to verify the nationality of every user in real time, the company had only one way to comply: switch the models off for everyone. So it did, that same evening. A capability that hundreds of millions could reach in the morning was dark by night.

Anthropic says it believes this is a misunderstanding — that the government’s concern rests on a narrow technique already available in other companies’ models, and that pulling a commercially deployed system over so small a finding is an overreaction. The original letter offered no rationale; in the days since, the Commerce Secretary has said publicly that he feared the models could be used by the military intelligence services of China, Russia, or other countries of concern. There is a lawsuit already running between the two parties, from an earlier clash. And as I write this — with the models still offline, negotiators from both sides now meeting almost daily, and a promised technical rebuttal from Anthropic not yet public — the story is very much still unfolding. Hold everything that follows loosely; the facts are fresh and contested, and I would rather be honest about that than pretend to a certainty no one yet has.

But the shape of the thing is already clear enough to learn from. And the lesson is not theone the headlines are reaching for.

Two Bad Futures, Visible in a Single Weekend

A while ago, in this series, I laid out four possible futures for the AI transition. Two of them were genuinely dangerous: a world where private corporations accumulate so much power that democratic government becomes irrelevant, and a world where the state seizes control of these systems and holds a kind of power over individual life that the twentieth century should have taught us to fear. Most of the time, those two dangers are abstractions — things to argue about in the future tense.

This weekend, both of them became visible in a single event. That is what makes it worth your time.

Look at the corporate-power side first. A private company, accountable first and foremost to its investors, built a capability so consequential that its mere existence is now a matter of national security. It decided what to build, how powerful to make it, when to release it, and to whom. Hundreds of millions of people came to depend on these systems, woven quietly into their work and their tools — and not one of those people had any say in any of it. The decisions were made in a boardroom, on a development timeline, according to a business strategy. That is corporate power over the conditions of ordinary life, exercised entirely outside any democratic process.

Now look at the state-power side, because it appeared in the very same event. With a single letter, on a Friday evening, the government switched that capability off — worldwide, for everyone, by fiat, with no public explanation at the time, no hearing, no transparent process, no warning. A technology that hundreds of millions could use was rendered inaccessible by executive directive in a matter of hours. Whatever you think of the specific decision — and there is a real case for it, which I will come to — the raw fact is striking: the state demonstrated that it can reach into the most powerful technology of the age and turn it off like a tap.

And it did so in a way that revealed something almost as unsettling as the power itself: a possible clumsiness about the very thing it was regulating. The order barred access by “any foreign national,” but there is no way to verify the nationality of every user of a global product in real time — so an order ostensibly aimed at foreigners took the tool away from everyone, including the American cybersecurity defenders who had been using it to protect systems. Critics across the political spectrum called the move “cartoonish” and “baffling”; one widely shared reaction said it read like something written by someone who does not understand how the internet works. Anthropic warned that if the same standard were applied across the industry, it would “essentially halt all new model deployments” from every frontier provider, since comparable capabilities exist in other public models. Set aside whether the government was right to be worried. The manner of it — fast, sweeping,and seemingly not fully thought through — is its own quiet warning about what concentrated state power over this technology can look like in practice.

Here is what I want you to notice. These are not two competing stories, one good and one bad. They are two faces of the same problem. On one side, enormous private power over your future, unelected and unaccountable. On the other, enormous state power over the same future, fast and opaque. And caught precisely between them — affected by their collision, consulted by neither — is everyone else. Is you.

The irony runs even deeper, and it is worth sitting with. Just two days before the government wielded this authority, Anthropic’s own chief executive published an essay arguing that the government should have exactly this kind of power — the legal authority to block or reverse the release of frontier AI models, like an aviation regulator grounding an unsafe plane. He asked for the state’s hand. Two days later, he got it, in a form he did not like and is now fighting. I do not say this to mock him; the impulse behind his argument may well be right. I say it because it captures the trap perfectly. Even the people building this technology cannot control the forces they are summoning. If they cannot, consider how much less control the rest of us have.

You Have No Seat at the Table

Strip away the technical detail and the corporate intrigue, and what this episode exposes isa chair. An empty one. The chair where the ordinary person — the one who will actually live in the world these decisions are building — should be sitting. It is not occupied by someone unworthy. It simply isn’t there.

Think about who was in the room this weekend. The company that built the technology. The government that regulates it. Their lawyers. Their security officials. Their strategists. Every party with a seat had enormous power and a direct stake. And every one of them was deciding the terms of something that reaches all the way down into the life of a freelancer in Colombia, a logistics manager in Ohio, a graphic designer in Lisbon, a small-business owner in Manila — none of whom were represented, consulted, or even informed until they read about it afterward, if they read about it at all.

This is the thing I most want you to take from a story that otherwise feels so remote. The remoteness is the message. You feel far from this because you are far from it — structurally, deliberately, by the design of how these decisions get made. The most powerful technology in human history is being governed by a small number of corporate and state actors, at extraordinary speed, behind closed doors, and the people whose livelihoods it is reshaping have no formal place in any of it. Not a vote. Not a voice. Often not even a clear view.

And that absence is not a glitch. It is the pattern. The same concentration of power that lets a company and a government fight over a frontier model this weekend is the concentration that is quietly deciding whether your kind of work survives the decade. The standoff over Fable 5 and the silence on your own inbox are the far and near ends of one single fact: a transformation of immense consequence is being run by very few, very fast, with the rest ofus as spectators to our own future.

And Then Someone Pulled Up a Chair

Here is where the story turned, and turned in a direction worth paying close attention to — because it shows the empty chair is not destined to stay empty.

Within two days of the shutdown, more than eighty people — cybersecurity chief executives, the heads of security organisations, the engineers who actually defend computer systems for a living — signed an open letter to the Commerce Secretary and the National Cyber Director, asking them to lift the directive. Their argument was concrete and grounded: the order had stripped a powerful defensive tool away from the very people who use it to protect hospitals, infrastructure, and businesses, and it had done so without a justification that withstood scrutiny. They were not lobbyists for Anthropic. They were practitioners with a stake, and they organised, quickly, to put that stake on the record.

I want to be careful not to overclaim what this is. Eighty security executives are not “the public” — they are themselves a relatively powerful and specialised group, and their letter is one pressure among many in a fight still dominated by giants. But notice what they did,because it is exactly the thing this essay has been circling. They saw a decision being made without the people it affected. And instead of accepting that the room was closed, they pulled a chair up to the table and insisted on being heard. That is what claiming a seat actually looks like, in its early and imperfect form. Not a vote handed down from above, but people with a legitimate stake refusing to be treated as spectators.

It is a small thing, and it may change nothing about this particular decision. But it is the right instinct, and it is contagious. If eighty security professionals can put their absence on the record within forty-eight hours, the question for the rest of us is not whether a seat can be claimed. It is whether we will bother to claim ours.

The Honest Part: There Is a Real Case on Every Side

I would be breaking my own rules if I handed you a villain here, so let me be clear that I am not.

There is a serious, legitimate argument for what the government did. Models that can autonomously hunt down hidden flaws in software are genuine dual-use tools — a shield in the right hands and a weapon in the wrong ones. A state that takes its security seriously has real reason to want hard limits on who can access such a capability, and “we discovered a way to strip the safeguards” is, on its face, exactly the sort of thing a security apparatus should worry about. If the government is right about the risk, acting fast and worrying about the paperwork later is defensible. Caution about catastrophic capabilities is not tyranny.

And there is an equally serious argument on the other side. The company says the concern is narrow, the technique is already available elsewhere, and the response — pulling a system used by hundreds of millions over a minor finding — is disproportionate. There is a live lawsuit and a history of bad blood between these two parties, which raises theuncomfortable possibility that national-security language is being stretched around what is partly a power struggle. A government that can switch off widely used technology overnight, with — at least initially — no public explanation and no transparent process, is exercising a power that should make anyone who values accountability uneasy, regardless of whether this particular use of it was justified.

It is worth pausing on one symptom of that opacity. In the vacuum left by the missing rationale, suspicion rushed in: online, people began speculating that a rival company’s large political donation might have greased the decision. I want to be clear that there is no evidence for this, and I am not repeating it as a claim — it is a rumour, and a platform built on honesty does not trade in rumours. But the fact that such a theory could take hold and spread so fast is itself the point. When power is exercised without a clear, public, technically grounded explanation, people do not conclude “there must be a good reason I can’t see.” They reach for the cynical story — and an opaque process gives them no way to be proven wrong. The damage of secrecy is not only what it hides. It is the suspicion it breeds in thespace where an explanation should have been.

Both of these are true at once. That is not fence-sitting; it is the actual situation. And notice what follows from holding both honestly: the ordinary person should want neither unchecked corporate power over this technology nor unchecked state power over it. The danger is not that the wrong giant won this round. The danger is that only giants are playing, and that whichever one prevails, the chair where you should be sitting stays empty.

That is the real story. Not Anthropic versus the government. Both of them, versus the absence of anyone who represents the rest of us.

Why This Belongs to You

You will be tempted to file this under “tech news” — interesting, distant, not yours. I am asking you to file it somewhere else.

File it under evidence. Evidence that the decisions shaping the transition you are living through are being made entirely without you, by powers you cannot see clearly and cannot influence, at a speed that leaves no room for objection. This weekend it was a model called Fable 5. Next time it will be a rule about what these systems are allowed to do in your industry, or who is permitted to build them, or how the staggering wealth they generate is shared — and you will not be in that room either, unless something changes.

The lesson is not to pick a side in this particular fight. It is to notice the empty chair, and to understand that the time to demand a place at the table is now — while this is still a news story you can watch from a safe distance, and not yet a decision that has quietly remade your life before you were told it was being discussed.

The most powerful technology of our age was switched off overnight, for everyone, by people you did not elect and a company you do not own, and you found out afterward. Sit with how little say you had in that. Then ask yourself how much say you have had in any of the rest of it.

That question — calm, persistent, refusing to be waved away — is where a seat at the table begins. We are not in the room yet. But the first step to getting in is to stop pretending we don’t belong there.

The train is moving. The people deciding its route met this weekend. None of us were invited. It is long past time we asked why.

Key Terms Explained

Export-control directive: A government order restricting who may access a particular technology, typically used for sensitive goods with national-security implications. Here, applied to AI models — barring access by any foreign national, which in practice forced a worldwide shutdown.

Dual-use technology: A capability that can serve both beneficial and harmful purposes — in this case, AI that finds software vulnerabilities, useful for defending systems and dangerous for attacking them. Dual-use is what makes governing these tools genuinely hard.

Supply-chain risk designation: A label, historically reserved for foreign adversaries, marking an entity as a national-security threat. Anthropic was given this designation earlier in 2026 after declining contract terms, and is challenging it in court — context for the strained relationship behind this weekend’s events.

The accountability gap: The space between the immense power held by corporate and state actors over AI, and the near-total absence of any formal role for the ordinary people that power affects. The central theme of this piece.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of artificial intelligence — for the people living through the transition, not the people building it. Written by Miguel Aparicio. This piece analyses a developing news story and takes no political side; it reasons toward the structural point and leaves the partisan conclusions to the reader.

Because this story is still unfolding, treat the specific facts as provisional. As of this update, the two models remain suspended; Anthropic and government officials are negotiating almost daily toward restoring access, with both sides’ principals reportedly set to cross pathsat the G7; and the detailed technical rebuttal Anthropic promised had not yet been published. Any of this may have changed by the time you read it. The structural lesson —about who decides, and who doesn’t — is the part that will outlast the news cycle.

Subscribe to The Human Transition on Substack, and share this with one person who thinks it doesn’t concern them.

Sources and further reading:

• Anthropic — “Statement on the US government directive to suspend access to Fable 5 and Mythos 5,” June 12, 2026

• CNBC, Fortune, Axios, Al Jazeera, Tom’s Hardware, Tom’s Guide, Quartz — coverage of the suspension, June 12–13, 2026

• The Globe and Mail / Reuters — “Anthropic, Trump officials working toward deal to restore Fable 5 and Mythos 5,” and Lutnick’s stated rationale, June 15–16, 2026

• MobiHealthNews — coverage of the June 14 open letter signed by 80+ cybersecurity leaders

• Dario Amodei — “Policy on the AI Exponential,” June 10, 2026

• NPR — “Pentagon labels AI company Anthropic a supply chain risk,” March 6, 2026 (background)

Subscribe now

If you have followed this series from the beginning, you have absorbed a great deal of diagnosis. The professions already being hollowed out. The retraining promise that does not add up. The political failure on both sides. The demographic trap. The unspeakable question about control. The collapse of provider identity. The anger economy and its threat to democracy. The fragility of the institutions of care. The food question.

At some point, a reasonable person stops and asks the only question that really matters to them personally: so what do I actually do?

This essay is an attempt to answer that — honestly, and with a warning attached. The warning is this: what follows is not a checklist, and it is not a solution. Anyone who offers you a clean, confident, step-by-step plan for navigating something this large, this unpredictable and this fast-moving is either fooling themselves or selling you something. What I can offer is a posture — a way of orienting yourself inside the transition that gives you the best chance of adapting as things change. And things will change, faster than this essay can anticipate, which is exactly why a posture is more useful than a plan.

This is a waypoint, not a destination. The series is not finished — the financial system, currencies, stores of value, personal security, the future of education all still lie ahead, and each will add to the picture. So treat what follows as the framework as it stands right now, built to be extended, exactly as your own approach should be.

Start With the Speed Problem, Because Everything Follows From It

To navigate the transition, you have to first understand the single most important thing about its shape: it moves faster than anyone's ability to respond to it, and that gap is permanent.

Think about the sequence. A technology evolves today. It is deployed and begins to disrupt tomorrow. The effects are felt by real people the day after. And the response — the policy, the protection, the adjustment that is supposed to catch the people who fall — does not come. Not because no one cares, though often no one does, but because the institutions responsible for responding operate on timescales that cannot possibly keep pace. Legislation takes years. Electoral cycles take years. International coordination takes longer still. The technology iterates in months. By the time a policy is debated, drafted, passed, and implemented, the thing it was meant to address has already changed shape three times.

This is not a temporary failure that better leadership will fix. It is structural. Deliberative institutions are, by design, slow — and that slowness, which is a virtue in stable times, becomes a fatal liability when the thing being deliberated about is moving at the speed of software deployment. There is no version of this where governments and policymakers catch up. The lag is built into the nature of the two systems.

The practical consequence is uncomfortable but clarifying: on the timescale that matters to your actual life, no one is coming to save you. Not the government, not the regulators, not the companies, not some future policy. This is not an argument against fighting for those structural responses — you should, and this series has argued relentlessly that you must. It is simply a recognition that you cannot wait for them. The structural fight operates on one timescale; your life operates on another. You have to navigate the second while pushing on the first.

That recognition is not despair. It is the premise that makes everything else necessary. Once you accept that no rescue is arriving on your timescale, the question stops being "when will this be fixed?" and becomes "how do I stay adaptive inside something that won't be?" That is a much better question, because it is one you can actually act on.

The First Posture: Integrate, Don't Resist — For Now

The first and most immediate orientation is the one that has run through this entire series: integrate the technology into whatever you do, rather than resisting it. Resistance — refusing to learn the tools, hoping the wave passes, treating AI as something to be avoided — is as futile as the fifteen-day retraining course, and for the same reason. It does not stop the wave. It just ensures you are swept up by it earlier, less prepared, and alone.

The same is true at every scale. The individual who integrates AI into their work is more productive and more durable, for now, than the one who refuses. The business that integrates outcompetes the one that does not. The country that integrates keeps pace; the one that tries to opt out simply falls behind and gets run over first. It is unrealistic to imagine that every government, or every company — Google, Anthropic, OpenAI, the rest — will agree to stop. They will not. The wave is coming regardless of what any individual does. Given that, refusing to engage is not principled resistance. It is unilateral disarmament.

But — and this is the part that matters most, the part that keeps integration from becoming surrender — integration is not endorsement, and it is not permanent. Using the tools does not mean approving of how the transition is being run, or accepting the concentration of power behind it. You can integrate AI into your work and remain its fiercest critic. And "for now" is doing essential work in that phrase, because the same forces that make integration smart today will, on a longer timescale, come for the integrated worker too. The AI-assisted professional of this year is on the same trajectory as everyone else; integration buys time, it does not confer immunity. So you integrate with open eyes: to stay viable in the near term, to buy yourself room to adapt, and to keep your footing while you push for the structural responses that individual adaptation can never replace. Integrate to survive the present. Never mistake it for having solved the future.

The Second Posture: Stay Agile, and Hold Every Plan Loosely

Because the ground keeps moving, no fixed strategy survives for long. The specific skill that is valuable this year may be automated next year. The safe-seeming role may not be safe in eighteen months. The plan that makes sense today is a hypothesis, not a guarantee.

This means the most valuable capacity is not having the right answer — it is the ability to keep updating. Agility itself is the skill. The person who navigates this best is not the one who picked the perfect strategy in 2026, but the one who stays alert, notices when the ground shifts, and adjusts without waiting for permission or certainty. Hold your plans firmly enough to act on them and loosely enough to abandon them when the evidence changes. Treat your own strategy the way this series treats its own framework: provisional, useful for now, and open to revision as reality develops.

This is uncomfortable, because human beings crave stability and a settled answer. But the demand for a settled answer is precisely the vulnerability that both the techno-optimists and the politicians exploit. They offer you certainty — abundance is coming, or this policy will fix it — because certainty is what frightened people want to buy. The healthier posture is to become comfortable with provisionality: to stop waiting for the moment when everything settles down, because in a fast-moving transition, that moment does not arrive. You learn to function well inside continuous change rather than enduring it while waiting for it to end.

The Third Posture: Keep Asking Why — and Stay Suspicious of Anyone Selling Certainty

Here is something this series insists on, and that I want to state as directly as I can: keep asking why this is happening, and refuse to let it proceed unexamined simply because it is presented as inevitable.

This matters because of a strange feature of the entire transition: it is the most consequential thing happening to ordinary people's lives, and almost no one was asked about it. No population voted for the restructuring of its economy. No worker consented to the obsolescence of their trade. The decisions that are reshaping the lives of billions are being made by a small number of companies and absorbed by a political system too slow to respond. In that situation, the act of questioning — persistently, stubbornly, out loud — is not negativity. It is one of the few forms of agency available. Why is this happening this way? Who decided? Who benefits? Who bears the cost? Why is no one being asked?

And from that questioning flows a necessary suspicion of the two groups who will spend the coming years telling you what to think about all this.

The first is the technology class — the founders, the investors, the public faces of the companies driving the transition. They will tell you a story of abundance: that the disruption is temporary, that the benefits will be shared, that what is coming is essentially a kind of liberation. Some of them may even believe it. But they have enormous financial and ideological incentives to keep you optimistic and passive, and as this series has documented, their private behaviour — the escape hatches, the hedges — does not always match their public reassurance. Be suspicious of the abundance story. Not dismissive — suspicious. Ask who distributes the abundance, and what happens to you in the meantime.

The second is the political class. They will tell you a different story: that they have it handled, that some bill or programme or reform will manage the transition, that you can trust the process. As this series has shown, on the timescale that matters they cannot deliver this, and many of them know it. Be suspicious of the politician's reassurance too — and equally suspicious of the politician who offers you a scapegoat instead of a solution, because as the essay on the anger economy argued, the redirection of your legitimate grievance toward a false target is the oldest move in the book and it changes nothing about the actual structure of power.

Suspicion of both is not cynicism, and it is not conspiracy thinking. It is the appropriate epistemic stance toward people who hold power and have a story to sell. Trust what survives scrutiny. Be wary of anything that asks you to stop scrutinising.

The Fourth Posture: Build Resilience From the Ground Up — and Build It With Others

All of the above is mindset. This is where it touches the ground.

The earlier essays in this series began assembling something that this framework now makes explicit: a resilience stack, built from the bottom up. At the base sits the most fundamental layer — food, water, the basic physical needs, the things the food essay argued are the foundation everything else rests on. Above that sits financial resilience — reducing fragility, lowering dependence on a single income, the material the financial essays still ahead will develop in depth. And running through all of it is the layer that the food essay and the sanctuaries essay both arrived at independently, and that I want to put at the centre of this framework: community.

Because here is the resolution to the hard truth I started with. Yes, on the institutional timescale, you are on your own — no rescue is coming fast enough. But "on your own" does not mean alone, and it must not become alone, because the lone individual is the most fragile unit there is. The food essay made this point starkly: the person who is the only one producing food in a collapsed community is not safe, but a target. The sanctuaries essay made it too: privatised, individual security fails at scale. Every strand of this series that has followed the logic of resilience to its end has arrived at the same place. The resilient unit is not the prepared individual. It is the connected community.

This is what self-reliance actually looks like when it is done intelligently rather than as a fantasy. Not the survivalist alone with his supplies, but the person embedded in a web of relationships — neighbours, family, local networks, people who produce different things and look out for one another — where security and capacity are shared rather than hoarded. No institution is coming to save you on the right timescale. That is precisely why you build, with the people around you, the resilience the institutions are failing to provide. The community is not a nice-to-have or a moral flourish. It is the single most practical thing in this entire framework, because it is the only form of resilience that does not collapse the moment one individual's plan fails.

You cannot navigate this alone. The good news buried in that hard sentence is that you are not supposed to.

Holding It Together

So the framework, as it stands at this waypoint, is a posture made of a few reinforcing parts.

Understand that the transition moves faster than any institution can respond, and that the lag is permanent — so stop waiting for rescue and start adapting now. Integrate the technology rather than resisting it, but never mistake integration for safety or endorsement — it buys time, for now. Stay agile, hold every plan loosely, and get comfortable functioning inside continuous change rather than waiting for a stability that will not come. Keep asking why this is happening and who it serves, and stay suspicious of everyone selling you certainty, whether they are promising abundance or promising they have it handled. And build resilience from the ground up — food, then finances, and above all community, because the connected group is the only unit of resilience that holds when individual plans fail.

None of this solves the transition. Nothing an individual does can, because the problem is structural and the solution must be too. But this posture lets you navigate it with open eyes, retain your agency, avoid the worst of the false promises, and keep your footing while the ground moves — all while continuing to push for the structural responses that are the only real answer.

That is enough to start with. It is not enough to finish with, and it is not meant to be. The series continues, and so does the framework — because in a transition this dynamic, the only approach that survives is one built, like the transition itself, to keep changing.

The train is moving. You cannot stop it, and you cannot step off. But you can learn to keep your feet under you while it moves — and you can make very sure you are not standing there alone.

Key Terms Explained

The response lag: The structural gap between the speed of technological deployment (months) and the speed of institutional response (years). Because it derives from the inherent slowness of deliberative institutions, it is permanent rather than a temporary failure — meaning no policy rescue arrives on the timescale of an individual's life.

Posture vs plan: A plan is a fixed sequence of actions that assumes a stable environment. A posture is a way of orienting and continuously adapting that assumes the environment will keep changing. In a fast-moving transition, a posture is more durable than a plan.

Integration without endorsement: Adopting AI tools to remain viable while continuing to criticise how the transition is being run. Integration is a survival strategy and a way to buy time — not approval of the system, and not permanent safety.

Provisionality: Treating one's own strategy as a working hypothesis open to revision, rather than a settled answer. The capacity to function comfortably amid continuous change, rather than enduring it while waiting for stability that will not arrive.

The resilience stack: The layered structure of resilience built from the bottom up — physical needs (food, water) at the base, financial resilience above, and community running through all of it as the binding layer. The connected community is the only unit of resilience that does not collapse when an individual plan fails.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

This is a waypoint in an ongoing series. Still ahead: the financial system, currencies and stores of value, personal and collective security, and the future of education — each of which will extend this framework further.

Subscribe to The Human Transition on Substack to follow the rest.

Subscribe now

Let me be transparent with you from the first line, because this essay touches a subject where hidden agendas are common and trust is everything.

I am vegan. I run an animal sanctuary. So when I think about food production, I am obviously not a neutral observer — I see plant-based agriculture as the right path, and you should know that before you read a word of what follows. I am telling you this deliberately, at the start, because the argument I am about to make does not depend on my values, and I want you to be able to check that for yourself. If this essay only works for people who already share my ethics, it has failed. The test of the argument is whether it holds for someone who eats meat, intends to keep eating meat, and cares only about navigating the transition ahead in the most rational way. I believe it does. Let me show you why, and you can judge.

Fundamentally, this is a resilience essay. It happens to reach a conclusion that aligns with my values, but it reaches that conclusion through economics and technology, not ethics — and that is the only reason it belongs in this series at all.

The Question This Series Forces

Everything this series has documented points toward a single practical question that thoughtful people are increasingly asking themselves: in a world where conventional income is becoming less reliable, where the economy is being restructured by forces no individual controls, what can a person or a community do to build genuine resilience?

Food is the most fundamental answer. Before income, before savings, before any abstract store of value, there is the basic question of whether you can eat. Throughout history, in every period of economic upheaval, the capacity to produce or reliably access food has been the most basic form of security there is. A person with a productive piece of land and the knowledge to use it has a kind of resilience that no purely financial asset can match.

So if the transition ahead is as disruptive as this series has argued — if income becomes unreliable, if supply chains become more fragile, if the systems we depend on come under strain — then food production becomes one of the most rational things to think seriously about. Not as a survivalist fantasy, but as a practical hedge, available at many scales: a household garden, a community plot, a smallholding, a regional food system.

The question is not whether food resilience matters. It is what kind of food production actually delivers resilience in the world that is arriving. And that is where the argument gets specific — and where, I think, the answer points clearly in one direction regardless of anyone's ethics.

Pillar One: The Efficiency Argument

The first pillar is the simplest and the least controversial, because it is a matter of physics and arithmetic rather than opinion.

Plant-based agriculture produces dramatically more usable food per unit of input — per acre of land, per litre of water, per unit of energy — than animal agriculture does. This is not a contested claim. It follows from the basic biology of the food chain. When you grow crops and eat them directly, you capture the food energy directly. When you grow crops, feed them to animals, and then eat the animals, you lose the large majority of that energy in the conversion. The animal spends most of the calories you feed it simply staying alive, moving, and maintaining its body heat. What is left as edible meat is a fraction of what went in.

The figures are stark and well established. Animal agriculture occupies a large share of the world's agricultural land while producing a much smaller share of its calories and protein. A significant proportion of global cropland is used not to feed people directly but to grow feed for livestock. Research consistently shows that the majority of agricultural land-use change — including deforestation — is driven by agricultural expansion, much of it for animal feed or grazing. In a world of stable abundance, this inefficiency is affordable.

In a world of tightening resources, fragile supply chains, and economic stress — the world this series has been describing — efficiency stops being an abstract virtue and becomes a survival advantage. A food system that produces more nutrition from less land, less water, and less energy is simply more resilient. It is less exposed to input costs, less vulnerable to supply disruption, and accessible to more people with fewer resources. If you are building food resilience from a position of economic constraint — which is the position the transition is pushing more and more people toward — the efficiency of plant-based cultivation is not an ethical preference. It is the practical path of least resistance.

This argument requires no veganism. It requires only the recognition that when resources are constrained, the more efficient system wins.

Pillar Two: The Disruption Argument

The second pillar is the one that connects this essay directly to the central thesis of the entire series, and it is the one most people have not yet thought through.

This series has argued, repeatedly, that AI and automation are disrupting industries by making older ways of doing things economically obsolete. The same logic is now arriving in food production — and it points directly at animal agriculture as one of the industries most exposed to technological disruption.

Consider what is already happening. Cultivated meat — real animal protein grown from cells, without the animal — has moved from laboratory curiosity to a fast-scaling industry. The broader cultured-meat ecosystem now includes well over a hundred companies globally, and, in October 2025, Believer Meats secured full USDA approval for a 10,000-ton-per year cultivated chicken facility, a $123 million plant designed for larger-scale market supply in 2026 and beyond. The cost curve has been dramatic: companies have brought production costs down by roughly 99 percent in less than a decade. AI is accelerating this directly — companies leveraging AI in cell cultivation are seeing up to a 40 percent reduction in production costs, and one firm achieved a 400 percent increase in output through large-scale bioreactor innovations.

The same is happening with dairy, and here is where the story becomes genuinely striking. There is a publicly traded Israeli company called Wilk — formerly Biomilk — that has developed and patented methods to produce real milk from cultured cells, without a cow. In early 2022 Wilk received US patent approval for its methods and systems for in-vitro milk production, making it an exclusive patent holder for technologies that produce real milk from cultured cells and even real human breast milk from cultured human mammary cells. And the detail that should make anyone sit up: Wilk is backed by Coca-Cola — it received a $2 million investment from Coca-Cola's Central Bottling Company arm, Coca-Cola Israel, which owns one of Israel's largest dairy processors. The company has already produced the world's first yogurt made with cell-cultured milk fat, validated by external laboratories as meeting the requirements to be called yogurt.

Pause on what that means. A global beverage corporation has invested in, and a company has patented, the capacity to produce real milk — chemically and biologically genuine dairy — in a factory, without any cow ever being involved. If that technology scales, milk could be manufactured the way other industrial food ingredients are: in bioreactors, at volume, by corporations, entirely decoupled from animals and the land and labour that dairy farming requires. The same approach is being pursued for the components of human breast milk, aimed at the infant formula market, which is projected to exceed $100 billion. This is not a fringe experiment. It is patented, corporate-backed, and already producing validated products.

Now apply the logic this series has applied to every other industry. When a technology can produce the same end product more cheaply, more controllably, and at industrial scale, the older, less efficient method does not survive on sentiment. It is disrupted. The same structural forces dismantling white-collar and blue-collar work are beginning to arrive in animal agriculture — and they come from two directions at once: cultivated and cell-based products replacing animal products at the high end, and the sheer economic efficiency of plant cultivation at the base.

I want to be scrupulously honest about the uncertainty here, because overstating it would betray the whole point of this platform. Cultivated meat is not yet at price parity, and serious analysts disagree about when it will be. McKinsey has noted that the industry would need many times the fermentation capacity that currently exists in the entire global pharmaceutical industry to meet projected demand, and estimates it will take until at least 2030 for these proteins to reach price parity. Some independent cost models put large scale cultivated meat far above conventional meat for the foreseeable future, and skeptics argue the science faces hurdles that optimistic timelines gloss over. The disruption is not guaranteed to arrive on any particular schedule, and a responsible analysis has to say so.

But the direction is not seriously in dispute, and the direction is what matters for a resilience decision. Whether animal-product disruption arrives in five years or fifteen, betting your food resilience on the long-term economic durability of conventional animal agriculture means betting against the same technological forces this entire series has shown to be reshaping every other industry. That is not a bet a clear-eyed person makes. The rational move, when an industry is visibly in the early stages of technological disruption, is not to build your future on it.

Pillar Three: The Income-Collapse Argument

The third pillar ties the food question directly to the economic thread running through this series.

Animal agriculture, at every scale, is capital-intensive and input-heavy. Livestock require feed, water, shelter, veterinary care, and significant land. The economics depend on a steady flow of inputs and a market of customers with the disposable income to pay a premium for animal products. Both of those conditions are threatened by exactly the dynamics this series has documented.

As labour income erodes across the population — the central thesis of this entire project — two things happen simultaneously. On the demand side, animal products, which are generally more expensive than plant-based staples, become harder for stressed households to afford; in every historical period of economic hardship, consumption shifts toward cheaper plant-based calories out of simple necessity. On the supply side, the capital and inputs required to sustain animal agriculture become harder to finance and maintain, especially for the small producer.

For an individual or community building food resilience under economic pressure, this points clearly in one direction. Plant cultivation requires less capital, fewer inputs, less infrastructure, and less specialised knowledge to begin. A person can start growing food on a balcony, a garden, a community plot, with minimal resources. The barrier to entry is low, the input costs are low, and the resilience benefit is immediate. Animal husbandry, by contrast, demands more capital, more space, more continuous input, and more risk — precisely the things that become scarce as economic security erodes.

So the income-collapse argument and the efficiency argument converge: when resources and income are constrained, plant cultivation is simply the more accessible, more affordable, and more defensible form of food resilience. Again — no ethics required. Just arithmetic applied to a constrained budget.

Where AI Helps Rather Than Harms

There is an important and hopeful counterpoint worth making here, because it connects to the "integrate, don't resist" principle that runs through this series.

In food production, AI and technology are not purely a threat — they are also a powerful tool for exactly the kind of small-scale resilience this essay is advocating. AI-driven tools are making small farms and even household-scale growing dramatically more productive: optimising irrigation, predicting weather and pest risks, improving soil management, identifying plant diseases early, and helping growers get far more yield from limited land with limited expertise. The same technology that disrupts conventional employment can make a small plant-based growing operation more viable than it has ever been.

This is the constructive face of the transition. A person building food resilience today has access to knowledge and tools — much of it AI-enabled, much of it free — that no previous generation of small growers had. Integrating that technology into a plant-based growing effort is the practical embodiment of everything this series argues: not resisting the technological wave, but using it to build the resilience that the same wave is making necessary. The smallholder who uses AI to optimise a plant-based plot is not fighting the future. They are navigating it intelligently.

Food as the Foundation — and an Honest Caveat About What It Is Worth

It is worth stepping back to place this in the larger structure that this series has been building, because food is not a standalone topic. It is the base layer of resilience. Everything else this series will examine in the way of resilience — savings, stores of value, crypto, gold, security, community — sits on top of a single assumption: that the most basic needs are met. Money is a claim on goods. A store of value is a bet that the claim will still be honoured later. All of it is abstraction stacked on the physical reality of whether there is food, water, and shelter. Food sits at the bottom of that stack, which is why this essay sits where it does. If the resilience argument has a foundation, this is it.

But I want to be honest about the nature of the claim I am making, because the entire credibility of this platform rests on not overstating things, and food-resilience arguments are unusually prone to romanticism.

What I have argued is grounded in historical evidence, and that is exactly what I want to flag: it is historical, not predictive. Across recorded history, in periods of economic collapse, war, and upheaval, those who could produce some form of food were, at minimum, able to eat — and usually able to offer something of value to others, which gave them a place in whatever order survived. That is a robust pattern. It is the reason the instinct toward food resilience is sound. But it is a pattern drawn from the past, and the transition this series describes will definitely not resemble the past. We do not know how food production will function in a future shaped by technologies and disruptions we have never lived through. Cultivated and manufactured food could make small scale growing economically irrelevant. Climate instability could make it precarious. The honest position is that producing food is a rational hedge based on what has reliably been true before — not a guarantee about what will be true next. I am offering you a well-founded assumption, clearly labelled as one, not a certainty dressed up as advice.

And there is a harder truth that the romantic version of this argument always omits, one that connects directly to a question this series will take up later. If you are the only person producing food in a community that has collapsed around you, you are not safe. You are a target. Resources that are scarce and visible attract those who lack them. The lone homesteader with the full pantry, surrounded by hungry neighbours, is not in a position of security — he is in a position of acute danger, and history is unambiguous about how that tends to end. This is the same lesson the essay on the billionaire escape hatch pointed toward: privatised, individual security fails at scale, because no individual can defend a valuable resource against a desperate enough surrounding world.

The conclusion that follows is important, and it reframes everything in this essay. Food resilience is not, ultimately, an individual strategy. It is a collective one. The grower is safe not when he has the only garden, but when he is embedded in a community where enough people produce, share, and depend on one another that food security is a shared condition rather than a private hoard. The resilient unit is not the prepper with the bunker. It is the community that feeds itself together. This is why, throughout this series, the argument keeps returning to the same place: the deepest form of resilience is not what you can stockpile alone, but the strength of the web of people around you. Food is the foundation of resilience — but community is the foundation of food.

What I Am and Am Not Saying

Let me close by being precise, because precision is what keeps this honest.

I am not telling you to become vegan. That is my choice, disclosed openly, and it is not the point of this essay. I am not making a moral argument about animals here — that is a different conversation, and this platform is not the place I am choosing to have it.

I am saying something narrower and, I think, harder to dispute: that if you are thinking about food as a form of resilience in the face of the transition this series has documented — and you should be — then the rational place to focus your effort is plant-based cultivation. Not because it is kind, though I believe it is. Because it is efficient, because it is accessible under economic constraint, and because it is on the right side of the technological disruption that is coming for animal agriculture just as it is coming for everything else.

A meat-eater who takes the transition seriously and follows the economics and the technology should reach the same practical conclusion about where to invest their food-resilience effort as I do — even while continuing to eat meat for as long as it remains available and affordable, which for many people it will for some time yet. The conclusion does not require my values. It only requires clear thinking about efficiency, disruption, and economic constraint. That is the whole argument, and it stands on its own.

There is a connection here to my sanctuary, and I will name it plainly rather than let it sit as an unstated bias. Running Namigni is what made me think hard about food systems in the first place, and yes, it shapes how I see this. But the resilience case I have laid out is one I would make, and would believe, even if I had never founded a sanctuary — because it is built from the same analysis of efficiency, technological disruption, and economic constraint that runs through every other essay in this series. I have simply applied that analysis to food. The conclusion it reaches happens to align with how I already live. I have tried to show you the reasoning so you can trust the conclusion regardless of whether you share the life.

The train is moving. It is coming for how we produce food just as surely as for how we earn a living. The most resilient response — for the vegan and the meat-eater alike — is to build food security on the foundation that the future favours, not the one the past was built on.

Key Terms Explained

Food resilience: The capacity to reliably produce or access food independent of fragile external systems. A fundamental form of security, available at scales from a household garden to a regional food system.

Feed conversion / trophic loss: The energy lost when crops are fed to animals rather than eaten directly. Because animals expend most consumed calories on living, only a fraction of the input energy is recovered as edible meat — making animal agriculture inherently less land- and energy-efficient than plant cultivation.

Cultivated (cell-based) meat: Real animal meat grown from cells in bioreactors, without raising or slaughtering an animal. A fast-scaling industry whose costs have fallen sharply, though it has not yet reached price parity with conventional meat.

Cultured / cell-based milk: Real milk produced from cultured mammary cells without a cow (or, for human milk, from human mammary cells). Patented by companies such as the Coca-Cola backed firm Wilk, and already used to produce validated dairy products.

Price parity: The point at which an alternative product costs the same as or less than the conventional one — the threshold at which mainstream adoption typically accelerates rapidly. For cultivated animal products, widely projected around or after 2030, though timelines are disputed.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio, who founded and runs Namigni Animal Sanctuary in Colombia. The author is vegan; this essay makes an economic and technological resilience argument rather than an ethical one, and discloses that starting point openly.

Subscribe to The Human Transition on Substack to receive the full series.

Subscribe now

Sources and further reading:

• Believer Meats / BioInformant — cultivated meat industry scaling and USDA facility approval, 2025–2026

• Good Food Institute — cultivated meat cost and price-parity research

• Cultivated Meat Shop / Forbes — AI-driven production cost reductions, 2025

• McKinsey — analysis of fermentation capacity constraints and price-parity timelines

• The Counter — "Lab-Grown Meat Is Supposed to Be Inevitable. The Science Tells a Different Story." (skeptical analysis)

• Trellis — "Lab Meat Has 3 Big Problems," updated February 2026

• The Times of Israel / VegNews / Food Dive — Wilk cultured milk, patent, and Coca-Cola Israel investment

• UN Food and Agriculture Organization — livestock emissions and agricultural land-use data

I should begin this essay by telling you where I am writing it from, because it is relevant to everything that follows.

I run and mostly self-fund an animal sanctuary. Namigni Animal Sanctuary, in Guasca, Colombia. I know, from the inside and in detail, what it costs to keep animals alive every day, and where the money to do it comes from. I know what it feels like to watch a donation base that is never quite enough, made up of people whose own circumstances are never quite secure. And I know — because I live it — that the financial fragility most sanctuaries operate under is not a failure of management. It is a structural condition of being a cause that cannot advocate for itself, funded by people whose ability to give is the first thing to disappear when the economy turns. And that includes my own ability to give, by the way.

This essay is about that structural condition, and about what the AI transition does to it. It is, on the surface, an essay about charities and sanctuaries. Underneath, it is an essay about something larger: the entire ecosystem of civil society — the charities, the shelters, the food banks, the cultural institutions, the community organisations, the sanctuaries — runs on a foundation of discretionary income. And discretionary income is precisely what the AI transition threatens. When people lose the security of their wages, the first thing they cut is not their rent or their food. It is the money they were giving to causes beyond themselves.

The question in the title is not rhetorical. If the people who fund sanctuaries and charities lose their incomes, who will feed the animals? Who will fund the shelters? Who will keep civil society alive? Nobody is asking this question. It deserves to be asked.

The Headline Numbers Look Fine. They Are Hiding Something.

If you look only at the top-line figures, the charitable sector appears healthy. Total charitable giving in the United States reached $592.5 billion in 2024, a new record high. Surveys of wealthy donors are reassuring: 93% of high-net-worth donors planned to maintain or increase their charitable giving in 2026, with nearly half planning to give more. Foundations are projected to increase their giving. On the surface, generosity is holding.

But the headline number is concealing something that should alarm anyone who cares about the long-term health of civil society. That record total masks a dangerous concentration in who is giving: for five consecutive years, the total number of Americans giving to charity has declined, dropping 4.5 percent in 2024 alone and accelerating a decades-long trend. The scale of the disappearance is staggering. Two-thirds of American households donated to charity in 2000; that number has fallen to about half — a loss of roughly 20 million giving households. And as the analysis pointedly noted, these millions of missing donors aren't billionaires.

This is the crucial structural fact. The money is increasingly coming from fewer, wealthier people, while the broad base of ordinary donors — the people who gave twenty, fifty, a hundred dollars because a cause moved them — is quietly eroding. Donor retention has fallen to just 31.9 percent, meaning fewer than a third of last year's donors give again. The drivers are precisely the ones this series has documented: concerns over inflation, job security, and student debt have made personal financial planning more cautious, reducing discretionary charitable spending — and a growing concentration of charitable dollars now comes from high-net-worth individuals while average and low-income donors give less or not at all.

Read that again, because it is the entire argument in miniature. The thing that makes ordinary people stop giving is economic insecurity about their jobs and their debts. And economic insecurity about jobs and debts is exactly what the AI transition produces, at scale, across the entire population of ordinary donors simultaneously.

The sector is already running on a narrowing base. The AI transition threatens to narrow it dramatically further — or perhaps even eroding it completely.

Why Discretionary Giving Is the First Thing to Go

To understand why this matters so much, it helps to understand the psychology and economics of charitable giving.

Charitable donations come, overwhelmingly, from discretionary income — the money left over after the essentials are paid for. Rent or mortgage, food, utilities, transport, debt payments: these come first. Giving comes from what remains. This means that charitable giving is one of the most economically sensitive forms of spending there is. When a household's financial security is threatened, giving is among the very first things to be cut, because it is, by definition, the spending that is easiest to stop. You cannot stop paying rent. You can stop the monthly donation to the animal sanctuary with a single click.

This is why the giving economy is so vulnerable to exactly the kind of disruption AI produces. The previous essays in this series have documented how AI erodes income — not always through outright job loss, but through falling wages, disappearing freelance work, reduced hours, and the pervasive insecurity of watching your industry contract. Every one of those mechanisms reduces discretionary income. And every reduction in discretionary income flows directly through to the causes that depend on it.

There is a second, subtler mechanism. Giving is not only about money. It is about a psychological state — a sense of security and surplus that allows a person to look beyond their own immediate needs and care about something larger. A person who is frightened about their own future, who does not know whether they will have an income next year, who is watching their savings erode, contracts their circle of concern. This is not selfishness. It is a natural and rational response to insecurity. But its effect on the giving economy is the same: as insecurity spreads, the psychological surplus that generosity depends on shrinks, even among people whose incomes have not yet fallen.

The AI transition, in other words, attacks charitable giving through both channels at once: it reduces the actual discretionary income available to give, and it erodes the sense of security that makes people willing to give what they have.

The Hierarchy of Causes — and Who Sits at the Bottom

Now we come to the part of this essay that is hardest to say plainly, but that needs saying.

Not all causes are equally funded, and they never have been. There is a hierarchy of charitable giving, and it is shaped by a brutal logic: causes that can make their beneficiaries visible, sympathetic, and articulate attract more support than causes whose beneficiaries cannot advocate for themselves. Children's charities, medical research, disaster relief, human welfare — these sit near the top, because the people they help can be seen, can tell their stories, can appear in the appeal and say thank you.

Animal causes sit lower. And within the animal world, there is a further hierarchy. Charismatic wildlife — the tiger, the elephant, the panda — attracts more support than farmed animals. Campaigning organisations that run advertisements and lobby for legislation attract more support than the unglamorous, day-to-day work of actually caring for animals. And at the very bottom of this hierarchy sit animal sanctuaries: the places that take in the actual animals, the rescued, the abandoned, the saved-from-slaughter, and then bear the relentless daily cost of feeding them, housing them, and caring for them for the rest of their natural lives.

A sanctuary cannot stop. A campaign can pause between fundraising drives. A research project can slow its spending. A sanctuary has living animals that must eat every single day, receive veterinary care, be sheltered through every season, for years. The cost is continuous, non negotiable, and invisible — there is no dramatic moment, no campaign victory, no headline. There is only the unending daily reality of keeping living creatures alive.

This is what I mean when I say the fragility is structural, not a failure of management. Sanctuaries operate at the intersection of the highest possible cost reliability — animals must be fed daily, forever — and the lowest possible funding security — they sit at the bottom of a giving hierarchy, dependent on the discretionary income of ordinary people who love animals. It is, financially, one of the hardest positions in the entire civil-society ecosystem. Most sanctuaries operate perpetually on the edge. I can tell you from direct experience that this is not an exaggeration.

Now apply the AI transition. The donors who support sanctuaries are, disproportionately, ordinary people of modest means who give because they care — exactly the broad base of small donors that the data shows is already eroding, and exactly the people whose discretionary income AI threatens most directly. The animals cannot advocate for themselves. They cannot retrain. They cannot adapt. They are entirely dependent on a chain of human generosity that the AI transition is pulling apart at precisely its weakest link.

If discretionary giving collapses, sanctuaries are not somewhere in the middle of the queue of affected institutions. They are first, because they were always last in line for the funding and first in line for the unavoidable cost.

This Is Not Only About Animals

I have focused on sanctuaries because I know them intimately and because they are the clearest illustration of the dynamic. But the argument generalises, and it should.

The same structural vulnerability applies across the entire ecosystem of civil society that depends on discretionary giving. The food bank that feeds the hungry. The shelter that houses the homeless. The hospice. The arts organisation. The community centre. The small charity working on a cause too unglamorous to attract foundation money. All of them run, to varying degrees, on the same foundation: the discretionary income and the psychological surplus of ordinary people who choose to give.

And here is the cruel compounding effect that makes this so dangerous. The AI transition does not only reduce the supply of charitable giving. It simultaneously increases the demand for what charities provide. As displacement spreads, more people need the food bank, the shelter, the support services, the safety net that civil society provides where the state does not. So at the precise moment when the income that funds these organisations is contracting, the need they exist to meet is expanding. The scissors close from both directions: less money coming in, more need walking through the door.

This is the systemic cascade that runs through this entire series, appearing here in one of its most poignant forms. The freelancer losing income, the family losing security, the displaced worker, the collapsing donor base, the overwhelmed charity, the sanctuary that cannot feed its animals — these are not separate problems. They are the same problem, expressed through every institution that depends on the economic security of ordinary people. When that security erodes, everything built on top of it begins to fail at once.

There is a further dimension worth naming, because it connects to the essay on democracy and radicalisation. Civil society organisations are not only providers of services. They are part of the social fabric that holds communities together, that gives people a way to contribute, belong, and find meaning beyond their work. When that fabric frays — when the organisations that connect people and channel their care collapse for lack of funding — the social isolation and loss of meaning that result feed directly into the political volatility documented earlier in this series. The collapse of civil society is not only a humanitarian loss. It removes one of the structures that makes communities resilient against the worst political consequences of the transition.

What Can Actually Be Done

This is not a counsel of despair, and there are honest things to be said about how to respond — though I will not pretend they are easy or sufficient.

For sanctuaries and charities themselves, the data points toward one clear lesson: the broad base of small, recurring donors matters more than ever, precisely because it is what is most at risk. Monthly recurring giving has become the most stable form of nonprofit revenue, now accounting for roughly 31 percent of online revenue and growing. An organisation funded by a wide base of committed monthly donors is more resilient than one dependent on a few large gifts or on the year-end surge. Building and holding that base — through genuine relationship rather than transactional appeals — is the single most important thing a small organisation can do to survive what is coming. This is not a marketing tactic. It is a survival strategy.

For those who care about a cause — and can still afford to give — there is a case for thinking about giving differently in this period. A modest, reliable, recurring commitment to an organisation you believe in is worth more to that organisation than a larger but unpredictable gift, because it allows planning and provides stability precisely when stability is scarcest. If you are someone whose income is, for now, secure, your steadiness becomes disproportionately valuable as others are forced to withdraw.

But I want to be honest about the limits of these individual responses, because honesty is the point of this entire platform. The fundamental problem is structural, and it cannot be solved by better fundraising or individual generosity alone. If the AI transition erodes the discretionary income of the broad population, no amount of clever donor cultivation will replace the lost base. The deeper solution is the same one this series keeps returning to: a societal response to the transition that preserves the economic security of ordinary people. A population with stable income gives. A frightened, displaced, insecure population cannot. The survival of civil society — of the sanctuaries, the food banks, the shelters, the whole web of organised human care — depends, ultimately, on whether the transition is managed in a way that preserves the broad based economic security that generosity grows from.

And there is a connection here to the agriculture and resilience argument that the next essay in this series will develop. Sanctuaries, like the smallholdings and community food systems that essay will discuss, are most resilient when they reduce their dependence on fragile external funding and build, where they can, toward greater self-sufficiency — growing feed, generating their own resources, embedding themselves in resilient local communities rather than depending entirely on a distant and fragile donation economy. This is not a complete answer either, and it is far more difficult to execute than it seems to be. But in a world where the external funding base is threatened, the capacity to depend a little less on it is itself a form of resilience.

The Question Remains

I began with a question, and I want to end with it, because I do not have a complete answer and I think it is more honest to leave it standing than to pretend otherwise.

If the AI transition erodes the discretionary income of ordinary people — the people who fund the sanctuaries, the shelters, the food banks, the whole ecosystem of organised human care (and I am included in this group both as a sanctuary founder and as a sanctuary funder) — who will keep these things alive? Who will feed the animals?

I run a sanctuary. I ask this question not as an abstraction but as something I confront in practical terms. And what I have come to believe is this: the survival of the causes that depend on human generosity is bound up completely with the broader question this series exists to address. There is no separate solution for sanctuaries, for charities, for civil society. Their fate is the fate of the economic security of the ordinary people who sustain them. If we navigate the transition in a way that preserves that security, the web of human care survives. If we do not, it will fray and tear at its weakest points first — and the animals, who cannot speak, cannot vote, and cannot advocate for themselves, will be among the very first to suffer the consequences of a transition they had no part in creating.

That is reason enough to insist the transition be managed well. Not only for ourselves. For everything and everyone that depends on us being secure enough to care.

The train is moving. And some of the passengers cannot hold on by themselves.

Key Terms Explained

Discretionary income: The money remaining after essential expenses (housing, food, debt, utilities) are paid. Charitable giving comes almost entirely from discretionary income, making it highly sensitive to economic insecurity.

Donor retention: The percentage of a charity's previous-year donors who give again. A key measure of fundraising health, currently low (around a third), making organisations dependent on constantly replacing lost donors.

The giving hierarchy: The pattern by which causes whose beneficiaries are visible, sympathetic, and able to advocate for themselves attract more funding than causes whose beneficiaries cannot. Animal sanctuaries sit near the bottom.

Recurring giving: Regular, automated donations (typically monthly). The most stable form of nonprofit revenue and the most resilient against economic shocks, because it is less subject to one-time decision-making.

The funding-demand scissors: The dynamic by which an economic crisis simultaneously reduces the funding available to charities and increases the need for their services — squeezing them from both directions at once.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it.

Written by Miguel Aparicio, who founded and runs Namigni Animal Sanctuary in Guasca, Colombia. The next essay examines agriculture and food as the most fundamental form of resilience — and makes an honest, practical case for why a particular kind of food system may be the most rational response to the transition ahead.

Subscribe to The Human Transition on Substack to receive the full series..

Subscribe now

Sources and further reading:

• Stanford Social Innovation Review — "The Collapse of Grassroots Giving Threatens Social Innovation," March 2026

• Giving USA Foundation — total charitable giving figures, 2024

• Foundation Source — 2026 Donor Survey

• Bank of America Study of Philanthropy — Charitable Giving by Affluent Households, 2025

• Fundraising Effectiveness Project — donor retention data

• BetterWorld — "2025 Nonprofit Roundup & 2026 Forecast," December 2025

• Marketplace — "Charitable Giving, a Year in Review," January 2026

• Baldwin CPAs — "Addressing the Decline in Charitable Giving and Volunteerism," 2025

Throughout this series, a single thread has surfaced again and again, in the essay on political failure, in the essay on families and identity, in the discussion of who absorbs the cost of the transition. It is time to address it directly, because it may be the most consequential consequence of the entire AI transition — more consequential, in the long run, than any individual job loss.

The thread is this: large-scale economic disruption, when it is not managed by competent institutions, does not stay economic. It becomes political. And historically, the politics it produces are not calm, deliberative, or constructive. They are angry, polarised, and dangerous to democracy itself.

This is not speculation. It is one of the better-established findings in political science, drawn from the study of every major economic disruption of the last century. And it is now being applied, by serious researchers, to the specific case of AI. The question this essay asks is the one that follows directly from everything this series has documented: if AI displaces workers at the scale the evidence suggests, and if political institutions respond as inadequately as they currently are, what happens to democracy itself?

The answer is sobering. But it is not hopeless — and the difference between the sobering outcome and the hopeful one turns out to depend on specific, identifiable choices that are still available to be made.

What History Already Knows

The relationship between economic insecurity and political radicalisation is not a new discovery. It is one of the most consistent patterns in modern political history, and it has been studied intensively since the economic shocks of the early twentieth century.

The core finding is straightforward: when significant numbers of people experience a sharp decline in their economic security and status, and when they perceive that established political institutions have no answer for them, support for radical and anti-establishment political movements rises. This was true in the 1930s. It was true after the deindustrialisation of the 1980s and 1990s. And it was true after the 2008 financial crisis, which preceded a surge in populist and far-right parties across the developed world.

Research on that most recent wave is precise about the mechanism. The economic insecurity caused by technology-driven automation is partly to blame for growing political polarization across the United States and Europe over the past decade — including the rise of the Alternative for Germany, Lega and Fratelli d'Italia in Italy, Vox in Spain, the Rassemblement National in France, and the Sweden Democrats — most of which significantly increased their vote share after the 2008 economic crisis.

The detail that matters most for the AI transition is what specifically drives the radicalisation. It is not simply poverty. Recent research indicates that support for the far right is not primarily coming from the poorest of the poor, but from the poorest quarter who feel they have something to lose — driven by relative deprivation, when people see their share of national income shrink while the top tiers grow, and by the latent threat of unemployment rather than unemployment itself. Strikingly, research shows that if just one person in a household works in a high-risk industry such as traditional manufacturing, the entire household becomes significantly more likely to shift toward radical-right parties.

This is precisely the psychological situation that AI displacement produces — and produces at scale. It is not only the people who have already lost their jobs. It is the far larger number who can see it coming, who feel the ground shifting beneath them, who have something to lose and watch it becoming less secure. The previous essay in this series described the collapse of provider identity and the anxiety rippling through families. This essay describes what that anxiety does when it is aggregated across millions of households and enters the political system.

Why the AI Transition Is Especially Dangerous

There is a specific reason to believe the political consequences of AI displacement could be more severe than those of previous economic disruptions, and it goes to the heart of what this series has been documenting.

The political scientist Jane Gingrich, in research on automation and politics, identified the crucial variable: the role of social compensation. State intervention after the 2008 crisis — such as unemployment benefits and dismissal regulations — substantially reduced the impact of unemployment on political polarization. Social programs very likely prevented much radicalization among voters in the face of economic shocks, while pronounced austerity and direct cuts to social programs may stimulate a backlash.

In other words: the radicalisation is not automatic. It depends on whether the institutions of society catch the people who fall. When they do — when there is meaningful unemployment insurance, retraining that actually works, a social safety net that preserves dignity and status during the transition — the political backlash is muted. When they do not, it is severe.

Gingrich also identified a reason the AI transition may be worse than globalisation. “Compensatory policies are less effective against the political pressures of automation than they were against globalisation, because workers displaced by automation face the risk of losing both income and status.” The job lost to a factory in another country can be framed as something taken by an external force. The job lost to a machine is something rendered obsolete — and with it, the worker's sense that their skills, their experience, their identity have any value at all. As the families and identity essay argued, this is a loss of status, not only income — and status loss is precisely what the research identifies as the more politically combustible factor.

Now combine this with everything this series has established. The displacement is large and accelerating. The freelance and informal economy, where much of the displacement is happening, is invisible to the statistics and excluded from the safety nets. The political class, on both left and right, is failing to respond at the necessary scale. And — critically — most governments did not set up large retraining systems in response to previous shocks, few have passed comprehensive AI legislation, safety-net expansions look unlikely at a time when many governments are laden with debt, and some governments are actively cutting social programs."

This is the precise combination of conditions that the historical research identifies as most dangerous: large-scale economic and status disruption, occurring simultaneously with the withdrawal rather than expansion of social compensation. It is, almost exactly, the recipe for the kind of political radicalisation that history teaches us to fear.

The Anger Economy

A group of political scientists writing in Foreign Affairs in late 2025 gave this dynamic a name: the anger economy. Their argument is that whatever the long-term employment consequences of AI turn out to be, it will soon become a major political issue — and if there is significant disruption, officials will be confronted by workers furious about jobs lost to machines, who will make their frustrations known at the ballot box.

The phrase captures something important. The political energy generated by AI displacement is not neutral. It is anger — and anger, in politics, is a resource that gets harnessed. The question is never simply whether the anger exists. The question is who captures it, and what they direct it toward.

This is where the asymmetry that this series identified in the essay on political failure becomes decisive. The two political traditions that stand to capture this anger will do fundamentally different things with it.

A revitalised left would, in principle, direct the anger toward the structural causes: the concentration of AI-generated wealth, the absence of redistribution, the failure to tax the productivity gains, the need for a genuine social response to the transition. Whether it has the courage and the freedom from donor capture to do this is, as the earlier essay argued, an open question. But the direction of its analysis at least points toward the actual problem.

The far right does something different and, historically, more immediately effective in moments of crisis. It captures the anger and redirects it — away from the structural causes, which are complex and difficult, and toward targets that are simple, visible, and emotionally satisfying. Immigrants. Minorities. Elites. Foreigners. The enemy within. This redirection has a crucial feature: it leaves the actual structure of power untouched. A politics that blames immigrants for AI-driven job losses does nothing to constrain the corporations building the AI, nothing to tax the wealth being concentrated, nothing to build the safety net that would actually protect people. It provides the emotional satisfaction of a target without the substance of a solution — which is precisely why it is so attractive to those who benefit from the existing structure of power remaining unchallenged.

The research on how this works in practice is striking. Recent analysis of European elections shows that far-right parties have optimised short-form video more effectively than any other political group, using platform-specific psychological techniques, trending sounds, and memes to de-demonise their image and target young voters — balancing crisis narratives of fear and uncertainty with aspirational narratives of hope and national pride. The anger economy is not only a political dynamic. It is a media dynamic, and the tools of AI-driven content production and algorithmic distribution are extraordinarily well suited to harnessing and amplifying it.

The Disinformation Multiplier

There is a second way the AI transition threatens democracy, and it compounds the first. The same technology that is displacing workers is also degrading the information environment in which democratic deliberation depends.

In 2026, widespread information disorder has become a destabilising systemic force with the potential to disrupt democracies and erode social cohesion, with AI-generated deepfakes becoming nearly indistinguishable from reality — making it harder to discern what is accurate and offering opportunists plausible deniability to deflect from what is true. The mechanism is corrosive in a particular way: “just knowing that deepfakes exist can make people doubt things they read and see — even the truth.”

This matters enormously in the context of the anger economy. A population that is economically insecure, angry, and losing faith in institutions is already vulnerable to political manipulation. Add an information environment in which truth and falsehood are increasingly indistinguishable, in which any inconvenient fact can be dismissed as a deepfake and any fabrication can be presented as evidence, and the conditions for democratic deliberation begin to break down. People cannot make good collective decisions if they cannot agree on what is real. AI is simultaneously generating the economic anger and degrading the shared factual foundation that a healthy democracy would need in order to channel that anger constructively.

The Backsliding Is Already Underway

This is not a future scenario. The democratic erosion that these dynamics produce is already measurable, and it is accelerating.

The V-Dem Institute's Democracy Report 2026 found that nearly a quarter of the world's countries were undergoing autocratization in 2025, including several in Europe and North America, with 44 countries becoming more autocratic and six of the ten newest cases in Europe and North America. The pattern, as one analysis described it, spreads less through ideology than through technique — like malware running across different operating systems — as populist leaders, once in office, weaken the institutions that might hold them accountable, reduce oversight, increase corruption, and produce more insecurity, in a cycle that repeats not because politicians are uniquely malicious, but because the incentives for moderation have collapsed.

That final point deserves emphasis, because it connects directly to the AI transition. The incentives for moderation collapse when the centre cannot deliver security and the extremes can offer either solutions or scapegoats. AI displacement, by generating mass insecurity that the moderate centre is failing to address, accelerates exactly this collapse of the incentives for moderation. It is not that AI directly causes authoritarianism. It is that AI generates the economic and social conditions in which authoritarianism becomes more attractive, more electable, and harder to reverse.

Whether It Has To Be This Way

This is the part that matters most, because the entire purpose of this series is to help people navigate the transition rather than merely fear it — and the historical research that delivers the warning also points toward the response.

The radicalisation is not inevitable. The single most important finding in the research is that social compensation works. The countries and periods in which economic disruption did not produce severe political radicalisation were those in which the institutions of society responded — with income support, with genuine retraining, with policies that preserved the dignity and status of displaced workers, with a visible commitment to ensuring that the costs of the transition were shared rather than dumped entirely on those least able to bear them.

This means the political danger of the AI transition is, fundamentally, a choice. Not an individual choice — a collective, political one. A society that responds to AI displacement with serious redistribution, a rebuilt safety net, genuine investment in the affected, and a credible commitment to sharing the gains of the transition will face a manageable political adjustment. A society that responds with austerity, denial, and the abandonment of the displaced will face the anger economy in its most dangerous form — and may not emerge from it with its democracy intact.

The tragedy is that we are currently, in most developed democracies, choosing the second path. In face of an overwhelming trend that displaces people in ways that they have not caused and cannot avoid, cutting social programs rather than expanding them is a reality. Failing to tax the extreme wealth being concentrated. Leaving the displaced to absorb the cost alone. We are, in other words, running the precise experiment that history has already told us produces the worst political outcomes.

But "currently" is not "permanently." The choice remains open. And this is where the individual reader is not powerless, even in the face of a structural problem. The single most useful thing a person who understands this dynamic can do is to refuse the redirection. To recognise, when the anger comes — and it will come — that the immigrant, the minority, the foreigner is not the cause of what is happening to them. It is not being stated here that massive immigration trends are not being inherently disruptive — they are. But they are not the cause of the main problems that average people are facing. The cause is structural: a transition being driven by concentrated corporate power, with no adequate social response, by a political class that has failed to act. Directing the anger at the scapegoat feels satisfying and changes nothing. Directing it at the actual structure of power — demanding the redistribution, the safety net, the democratic governance of AI that would actually address the problem — is harder, less immediately satisfying, and the only thing that has ever worked.

A democracy is, in the end, its citizens' collective capacity to direct their legitimate grievances toward real solutions rather than false targets. The AI transition will test that capacity as severely as anything has in a century. Whether democracy — or the American political concept of a republic — survives the test depends, in large part, on whether enough people can see clearly what is actually happening to them — which is the entire purpose of this work.

The train is moving. The anger is coming. The question is whether we aim it at the wall that needs to come down, or at each other.

Key Terms Explained

The anger economy: A term coined by political scientists in 2025 to describe the political energy generated by AI-driven economic disruption — and the competition among political movements to capture and direct it.

Relative deprivation: The experience of being deprived relative to one's expectations or to others, rather than in absolute terms. Research identifies it as a stronger driver of political radicalisation than absolute poverty.

Status loss vs income loss: The distinction between losing earnings and losing social standing and identity. Research suggests status loss is more politically combustible, and that automation (unlike globalisation) inflicts both simultaneously.

Social compensation: The set of policies — unemployment insurance, retraining, dismissal protections, income support — that cushion workers against economic shocks. Research consistently finds that robust social compensation substantially reduces the political radicalisation that economic disruption would otherwise produce.

Democratic backsliding (autocratization): The gradual erosion of democratic institutions and norms, often by elected leaders who weaken the checks that would hold them accountable. Measured globally by institutions such as V-Dem, and accelerating in 2025–2026.

The redirection: The political technique of channelling economic anger away from its structural causes and toward scapegoats — leaving the actual structure of power untouched while providing emotional satisfaction.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio. This essay analyses political dynamics structurally and does not endorse any political party or candidate.

Subscribe to The Human Transition on Substack to receive the full series directly.

Subscribe now

Sources and further reading:

• Brookings Institution — "Does Automation Increase Support for the Far Right?"

• Magistro, Borwein, Alvarez, Bonikowski & Loewen — "The Coming AI Backlash: How the Anger Economy Will Supercharge Populism," Foreign Affairs, October 2025

• Vlandas & Halikiopoulou — "Welfare State Policies and Far Right Party Support," West European Politics, 2022

• Jane Gingrich — research on automation, social compensation, and political backlash

• V-Dem Institute — Democracy Report 2026

• Carnegie Endowment — "AI and Democracy: Mapping the Intersections," January 2026

• CEPA — "Europe's Democratic Backsliding Is Spreading Like Malware," March 2026

• World Economic Forum — "How Cognitive Manipulation and AI Will Shape Disinformation in 2026," March 2026

• "Understanding the Rise of the Far Right," 2026 analysis of European elections

When we talk about AI and jobs, we talk about numbers. Unemployment rates. Displacement percentages. Wage curves. Contribution bases. The entire public conversation about the AI transition is conducted in the language of economics, as though the only thing at stake when a person loses their work is their income.

The income matters. The previous essays in this series have documented, in detail, how it is being eroded and why. But there is a second loss that runs alongside the first, and it is almost entirely absent from the public conversation because it does not appear in any economic statistic. It is the loss of identity. The loss of the answer to the most basic question a person asks themselves: who am I, and what am I for?

For most adults in modern societies, the honest answer to that question is bound up, more tightly than we like to admit, with work. "I am a teacher." "I am an electrician." "I am an accountant." "I run a small business." These are not merely descriptions of how a person earns money. They are descriptions of who that person is — to themselves, to their families, to their communities. When the work goes, the income is only the first thing to disappear. The self goes with it.

This essay is about that second loss. It is about what AI displacement does to identity, to the meaning of being a provider, and to the relationship between parents and children when the economic narrative that organised family life for generations stops making sense. It is, in some ways, the most important essay in this series, because it addresses the part of the transition that money cannot fix — and that no universal basic income, however generous, will fully resolve.

Work Is Not Only About Money

There is a body of research, developed over nearly a century, that explains why losing work is so devastating in ways that go far beyond the loss of income. The foundational insight comes from the psychologist Marie Jahoda, who studied unemployment during the Great Depression and developed what is now called the privation model.

Jahoda's insight was this: employment provides a paycheck, but it also provides a set of things that human beings need and that are easy to overlook precisely because, when you have a job, you receive them automatically. Work structures time — it gives shape to the day, the week, the year. It provides social contact beyond the family. It connects the individual to goals and purposes larger than themselves. It confers status and identity. And it requires regular activity. These are what Jahoda called the latent functions of work — the things employment provides beyond the wage.

The psychological damage of unemployment, Jahoda argued, is not explained only by the loss of money. It is explained primarily by the loss of these latent functions. A person who loses their job loses their time structure, their social world, their sense of purpose, their status, and a central part of their identity — all at once, and often without realising that this is what is happening to them. They know they feel adrift, worthless, disoriented. They do not always connect it to the fact that work was quietly providing the scaffolding of their entire psychological existence.

This is why the research on unemployment and mental health is so consistent and so grim. Job loss is among the most stressful life events a person can experience, ranking alongside bereavement and divorce. Unemployment significantly contributes to depression by causing a loss of identity, social relationships, and financial security. The relationship runs in both directions: poor mental health is both a cause and a consequence of unemployment, creating a downward spiral that becomes harder to escape the longer it continues.

Now consider what this means in the context of AI displacement. The conventional assumption — the one underlying every "retraining" programme and every proposal for income support — is that the problem of displacement is fundamentally a problem of money. Replace the income, the logic goes, and the problem is solved. But Jahoda's framework reveals why this is inadequate. Even a person whose income is fully replaced by a universal basic income would still lose the time structure, the social contact, the sense of purpose, the status, and the identity that work provided. The money is necessary. It is not sufficient. A society that solves the income problem of AI displacement while ignoring the identity problem will have solved only half of the crisis — and not the harder half.

The Specific Crisis of the Provider

Within this general picture, there is a particular case that deserves specific attention, because it is both especially acute and especially neglected in polite conversation: the crisis of identity among those — disproportionately, though not exclusively, men in their forties and fifties — whose sense of self is built around the role of provider.

The research is consistent on this point, even as the surrounding culture has shifted. “Studies have found that unemployed men experience higher mental health problems than women, with one tentative explanation being that men may perceive they have more responsibility for the financial running of the household” — a perception that, whatever one thinks of its origins, remains psychologically powerful for a great many people. “In studies of unemployment, men often used the stereotype of the "male breadwinner" to define who they were.” When that role becomes impossible to fulfil, the identity built upon it does not simply adjust. It can collapse.

This matters enormously for the AI transition, because the people most exposed to displacement in its current wave include precisely those whose identities are most likely to be organised around the provider role: mid-career workers, often with families, often with mortgages, often having built their entire adult identity around being the person who provides. When AI eliminates their role — not through any failure of their own, but through a structural change they did not cause and cannot control — the loss is not only financial. It is a loss of the thing that, for many of them, made them who they are.

The danger here is not only individual. It is social and political. A large population of people — disproportionately men, disproportionately mid-career, disproportionately in communities already under economic pressure — who have lost not only their income but their sense of purpose and identity, is one of the most reliably combustible social formations in modern history. The research on the political consequences of this kind of displacement is sobering. People who have lost their economic identity and feel that the institutions of society have abandoned them do not, in general, respond with calm civic engagement. They respond with anger, with a search for someone to blame, and with susceptibility to political movements that offer them a target for that anger. This is the thread that connects this essay to the political radicalisation that the earlier essay in this series identified: the collapse of provider identity is not only a mental health crisis. It is a precondition for political volatility.

None of this is an argument for nostalgia about traditional gender roles, or for pretending that the provider identity is the only valid one. It is simply an honest acknowledgment that a great many real people have built their lives around that identity, that AI is now removing the economic basis for it at scale, and that the consequences — psychological, social, and political — are serious and almost entirely unaddressed.

The Intergenerational Fracture: When "Do What I Did" Stops Working

There is a third dimension to this, and it may be the most poignant. It concerns the relationship between parents and children, and the quiet collapse of one of the oldest functions of parenthood: the transmission of a path.

For most of human history, parents have prepared their children for the future in large part by example and by guidance rooted in their own experience. "I built a life doing this. Here is what I learned. Do something like what I did, and you will be alright." The specifics changed across generations — the farmer's child might become a factory worker, the factory worker's child might go to university and become a professional — but the underlying structure held. Parents possessed hard-won knowledge about how the world worked and how to build a viable life within it, and they passed that knowledge to their children. It was one of the central things parents were for.

The AI transition breaks this. For the first time in modern history, a large proportion of parents genuinely do not know what to tell their children — not because they lack wisdom or care, but because the world their children will enter is so different from the one they themselves navigated that their hard-won experience may no longer apply. The professional who built a successful career over twenty-five years cannot confidently tell their child to follow the same path, because that path may not exist by the time the child reaches it. The knowledge that parents have always transmitted — this is how you build a life — has, for many, become uncertain at exactly the moment their children need it most.

The data captures the anxiety this produces. A national survey found that 53% of parents are concerned AI will narrow their children's job prospects, with many describing the AI driven future in words like "concerned," "cautious," "uncertain," and "worried.” 62% of parents reported discussing AI and the future of work with their teenagers within just the previous two weeks. 88% of parents said knowledge of AI will be crucial for their child's future, yet most were not sure whether schools are teaching it. Parents are, increasingly, abandoning the old script — questioning whether a four-year college degree is still the default path to success, and placing growing confidence in adaptability, critical thinking, and practical skills over credentials alone.

What is striking — and worth sitting with — is what the people building AI tell their own children. When the Wall Street Journal interviewed five AI leaders about what their own kids should study, their advice was not "learn to code." It was to move away from hyper specialization entirely. The reasoning is revealing: in an AI-saturated economy, accountability becomes the premium product — any job where a human must be legally, ethically, or physically responsible for an outcome is structurally insulated from disruption, because you cannot offshore accountability to an algorithm. The narrow specialist, who for thirty years was told that depth in a niche was the path to security, is now precisely the person most exposed. The architects of the transition know this, and they are advising their own children accordingly.

This is useful information, and the essay later in this series on personal navigation will develop it. But notice what it reveals about the intergenerational fracture. The old advice — specialise, get the credential, climb the ladder — is being abandoned even by the people who understand the system best. Parents who give their children the advice that worked for them may be preparing them for a world that no longer exists. And parents who recognise this often find themselves in the painful position of having no confident alternative to offer — only uncertainty, and the honest admission that they do not know.

There is a particular grief in this that deserves to be named. Part of what it means to be a parent is to feel that you can protect and prepare your child. When the future becomes genuinely unpredictable, when your own experience no longer reliably applies, that sense of protective competence is undermined. Many parents are experiencing a quiet version of this grief right now, often without naming it: the recognition that they cannot give their children the one thing parents have always tried to give — a reliable map of how to build a good life.

What Can Actually Be Said

This essay has deliberately resisted false comfort, because the losses it describes are real and the people experiencing them deserve honesty rather than reassurance. But honesty does not mean despair, and there are things that can be said that are both true and useful.

The first is that the loss of professional identity, however painful, is not the loss of the self. Work has carried so much of our identity precisely because, in the modern era, it absorbed functions — purpose, structure, community, status — that human beings can find in other places. Those functions predate the modern economy and will outlast it. Purpose can come from caregiving, creativity, community, craft, learning, contribution. Structure can be built deliberately rather than received automatically from a job. Social contact can be cultivated. Status — the real kind, the kind rooted in being valued by people who know you — was never actually dependent on a job title, even if it often felt that way. The task, for a person whose work has been taken by forces beyond their control, is not to replace the job. It is to rebuild, deliberately and consciously, the latent functions the job was quietly providing. This is hard. It is also possible, and people have done it through every economic upheaval in history.

The second is that the provider identity, while real and worthy of respect, is not the only way to be valuable to the people you love. A parent who can no longer provide financially in the way they once did is not thereby worthless to their family. The presence, the guidance, the love, the example of how to face adversity with dignity — these are forms of provision too, and in the long run they may matter more than income. This is not a platitude. It is a reframing that has allowed real people to survive real economic catastrophe with their relationships and their sense of self intact.

The third is for parents facing the intergenerational fracture. The honest admission "I don't know exactly what the future holds for you" is not a failure of parenting. It may be the most truthful and useful thing a parent can say, because it is accurate, and because it models the single most important capacity their children will need: the ability to face uncertainty without pretending it is not there. What parents can still transmit — what they have always been able to transmit, underneath the specific career advice — is character. Adaptability. The capacity to learn. Resilience in the face of disruption. Integrity. The ability to form relationships and contribute to a community. These are the things the AI leaders are quietly telling their own children to develop, and they are things every parent can model regardless of what happens to their own career.

The fourth, and the one this entire series keeps returning to, is that none of this is purely an individual problem to be solved through individual resilience. The scale of the identity crisis that AI displacement will produce is a collective challenge that demands a collective response — not only income support, but a serious societal conversation about what gives life meaning and structure in a world where traditional employment can no longer be assumed to provide it. That conversation has barely begun. It needs to.

The Loss That Money Cannot Fix

The central argument of this essay is simple, and it cuts against the grain of almost everything else being said about the AI transition. The displacement crisis is not only an economic crisis. It is an identity crisis. And the identity crisis is, in many ways, the harder of the two to solve, because it cannot be addressed by redistribution alone.

A universal basic income, if it were ever implemented at adequate scale, would address the income problem. It would not, by itself, address the loss of purpose, structure, status, and identity that the disappearance of work produces. It would not tell the displaced accountant who he is now. It would not give the laid-off engineer a reason to get up in the morning. It would not repair the parent's painful inability to offer their child a reliable map of the future. These are problems of meaning, not money — and a society that imagines it can navigate the AI transition by solving only the money problem is preparing to be surprised by the depth of the meaning problem.

The people building this transition talk about abundance — a future in which machines do the work and humans are freed to pursue creativity, leisure, and meaning. Perhaps. But meaning is not automatically distributed alongside material goods. It has to be cultivated, individually and collectively, and the institutions and traditions that once cultivated it — work foremost among them — are precisely what the transition is dismantling. The question of where human beings will find purpose, structure, and identity in a post-work world is not a soft or secondary question. It may turn out to be the central question of the entire transition.

The train is moving. It is carrying away not only our jobs, but a great deal of what we built our sense of self upon. The income, perhaps, can be replaced. The harder work — the work of rebuilding meaning, identity, and the bonds between generations on new foundations — is work that no machine can do for us, and that no policy alone can accomplish. It is, in the end, ours.

Key Terms Explained

Latent functions of work (Jahoda's privation model): The benefits employment provides beyond income — time structure, social contact, collective purpose, status, identity, and regular activity. The psychological damage of unemployment is explained largely by the loss of these functions, not only the loss of the wage.

Provider identity: A sense of self built primarily around the role of financially supporting one's family. Historically strongest among, though not exclusive to, men; psychologically powerful and especially vulnerable to the kind of structural displacement AI produces.

Intergenerational transmission: The passing of knowledge, values, and life-paths from parents to children. The AI transition disrupts this by making parents' hard-won experience potentially inapplicable to the world their children will face.

Accountability premium: The emerging principle that roles requiring a human to be legally, ethically, or physically responsible for an outcome are structurally resistant to automation, because responsibility cannot be delegated to an algorithm. Increasingly cited by AI leaders as the basis for career advice to their own children.

Meaning problem vs money problem: The distinction between the economic dimension of displacement (loss of income, addressable by redistribution) and the existential dimension (loss of purpose, structure, and identity, not addressable by money alone).

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

This is a sensitive topic. If you or someone you know is struggling with the psychological effects of job loss or a loss of sense of purpose, reaching out to a doctor, a mental health professional, or a trusted person can genuinely help — these are real challenges, and support makes a difference.

Subscribe to The Human Transition on Substack to receive the full series directly.

Subscribe now

Sources and further reading:

• Marie Jahoda — the privation model / latent functions of work (foundational research on unemployment and psychological wellbeing)

• Dawn R. Norris — Job Loss, Identity, and Mental Health (Rutgers University Press)

• "Unemployment and Mental Health: A Global Study," PMC, 2024

• Strandh, Hammarström et al. — gender differences in unemployment and mental health

• Penn Foster — "Parents of Gen Z Are Rethinking the Path to Success in an AI-Driven Job Market," January 2026

• Computer Weekly — "AI Leads Parents to Change Their Careers Advice to Children," November 2025

• Wall Street Journal — "What AI Executives Tell Their Own Kids About the Jobs of the Future"

• Vincent Hunt — "What to Tell Your Kids About Work in the Age of AI," 2025

This week, the New York Times reported that Peter Thiel — PayPal co-founder, Palantir chairman, one of the most influential investors and political donors in the technology world — has temporarily relocated his family to Argentina. He has reportedly purchased a $12 million mansion in one of the most exclusive neighbourhoods of Buenos Aires, enrolled his children in a local school, and met repeatedly with Argentine President Javier Milei, Economy Minister Luis Caputo, and Deregulation Minister Federico Sturzenegger.

The story went viral within hours. Most of the coverage focused on two angles: the tax angle — a proposed California wealth tax of 5 percent on assets over $1 billion could cost Thiel approximately $1.4 billion of his roughly $28 billion fortune — and the political angle, his ideological alignment with Milei's anti-tax, anti-regulation, anti-socialist government.

Both angles are real. Neither is the most important thing about this story.

Buried in the reporting is a detail that should arrest anyone who has been paying attention to what the people at the centre of the AI transition actually believe about where the world is heading. According to the New York Times, one of the considerations behind Argentina's appeal is its geography: in the event of a nuclear war that devastates all of North America, Argentina "would hardly be affected at all." Sources familiar with Thiel's thinking describe Argentina as a potential hedge against geopolitical risk — a "Plan B" if things go badly wrong in the United States.

This is the part of the story worth thinking about carefully. Not the tax. Not the politics. The escape hatch.

What the Best-Informed People Are Doing With Their Money

Peter Thiel is not an ordinary billionaire. He is one of the most significant figures in the infrastructure of the AI transition. Palantir, which he chairs, is among the most important AI and data companies in the world, deeply embedded in defence, intelligence, and government systems. He was an early investor in numerous AI and technology ventures. He has spent decades thinking, writing, and speaking about technological acceleration, civilisational risk, and the future of human society. Whatever one thinks of his politics — and they are highly contentious — he is unquestionably one of the best-informed people on earth about the trajectory of the technologies reshaping the global economy.

When a person with that level of insight and that level of resources begins building a physical escape route to one of the most geographically isolated, geopolitically insulated places on the planet, it is worth asking what they are insuring against.

Thiel is not alone in this behaviour, and that is the point. The phenomenon of ultra-wealthy technology figures acquiring remote properties, citizenship in multiple countries, and elaborate contingency plans has been documented for years. New Zealand became a notable destination for Silicon Valley bug-out properties over the last decade. Thiel himself already holds citizenship in the United States, Germany, and New Zealand, and has reportedly been pursuing a Maltese passport. The acquisition of a fortress in Buenos Aires — a 17,200-square-foot mansion, with residents reportedly already noticing increased private security activity around the property — fits a long-established pattern.

It is essential to be precise here, because this is exactly the kind of story that invites overstatement. There is no public confirmation that Thiel has permanently relocated or renounced his US residency. Reports indicate his Argentine home is, so far, his only known investment in the country. Newsweek's analysis of the available data concluded that while Thiel's move has sparked debate over a "billionaire exodus," the evidence does not suggest this is happening on a large scale — at least not yet. This is, most likely, an early-stage hedge rather than a full departure.

But the significance of the story does not depend on whether a wave of billionaires follows him. The significance is in what the behaviour reveals about the worldview of the people building the future the rest of us will live in.

The Pattern That Connects the Bunker to the Boardroom

There is a profound and underexamined contradiction at the heart of the technology elite's relationship to the future.

The dominant public narrative from the leaders of the AI industry is one of optimism. Abundance is coming. Disease will be cured. Energy will be nearly free. Humanity will be liberated from drudgery. The future is bright, and they are building it. This is the story told at conferences, in interviews, in the manifestos that circulate among the powerful.

And yet the private behaviour of some of the same people — and people adjacent to them — tells a different story. You do not build an escape hatch to a nuclear-insulated compound in the Southern Hemisphere if you are confident that the future you are building is bright. You build it because you understand, better than most, that the transition ahead carries genuine catastrophic risk — economic, social, geopolitical — and you have the resources to insure yourself against the downside while continuing to profit from the upside.

This is not necessarily hypocrisy. It may be entirely rational. If you genuinely believe there is, say, a 10 to 20 percent chance of catastrophic outcomes from the technological and geopolitical trajectory of the next two decades — which, as the previous essay in this series documented, is within the range that serious AI researchers themselves assign to existential risk — then building a hedge against that scenario is simply prudent risk management for someone who can afford it.

The problem is not that the wealthy are insuring themselves. The problem is what it reveals: that the people with the most information and the most resources are quietly preparing for the possibility that the transition goes very badly — while the public conversation they shape continues to reassure everyone else that there is nothing fundamental to worry about.

The billionaire buys a compound in Buenos Aires. He has the means to convert his understanding of risk into personal security. The teacher, the freelancer, the factory worker, the pensioner — the people this series has been documenting throughout — have no such option. They are told the future is bright, encouraged to keep consuming and planning as though the ground beneath them is stable, while the people who know better build their exits.

Why This Is Not a Reason for Paranoia — But Is a Reason for Attention

It would be easy to read this story as confirmation of a conspiracy: the elites know something catastrophic is coming, and they are fleeing while keeping the rest of us in the dark. That reading is both unfair and unhelpful.

The more accurate and more useful interpretation is this: there is no secret. The information is public. Geoffrey Hinton's warnings are public. The Statement on AI Risk is public. The economic data on displacement is public. The demographic projections are public. The people building escape hatches are not acting on secret knowledge. They are acting on public knowledge that they take seriously and that most people — encouraged by the reassuring narratives of politics and media — do not.

The lesson is not that catastrophe is certain. It is not. The lesson is that the impulse toward resilience — toward insulating yourself and the people you are responsible for against the possibility that the transition goes badly — is rational, and that the wealthiest and best-informed people on earth are acting on it.

The difference between Thiel and an ordinary person is not insight. The information is available to everyone. The difference is resources. Thiel can buy a $12 million compound in a nuclear insulated country. The ordinary person cannot. But the underlying logic — reduce your dependence on fragile systems, build resilience against disruption, do not assume the ground is stable simply because you are being told it is — is available at every scale.

For most people, that does not mean a bunker in Argentina. It means something far more modest and far more practical: reducing debt where possible, building skills and capacities that do not depend on a single fragile employer or industry, strengthening local and community resilience, maintaining the ability to produce or access essentials independently of distant and fragile supply chains, and refusing the false comfort that everything will simply continue as it has. These are not paranoid behaviours. They are the same behaviours the billionaires are exhibiting, scaled to the resources of an ordinary life.

The Question the Story Leaves Us With

Peter Thiel's move to Argentina will be debated for its politics, its tax implications, and its symbolism. But the question it should leave us with is simpler and more personal.

If the people with the most information about where the world is heading are quietly building exits, what does that tell you about the confidence you should place in the reassurances offered to everyone else?

Not that catastrophe is certain. Not that you should panic. But that the gap between the public narrative of inevitable abundance and the private behaviour of the people who shape that narrative is real, it is significant, and it is worth paying attention to.

The billionaires are buying compounds. The question for the rest of us is not how to follow them to Argentina. It is how to build, at the scale of an ordinary life, the resilience that their behaviour suggests even they believe is necessary.

Once again, the train is moving. Some passengers have already located the emergency exits. It is worth noticing where they are looking.

Key Terms Explained

Bug-out property: A secondary residence, typically in a remote or geopolitically stable location, acquired as a contingency against catastrophic disruption — economic collapse, war, civil unrest, or environmental disaster. Increasingly common among ultra-high-net-worth individuals over the past decade.

Geopolitical hedge: An investment or relocation undertaken not primarily for financial return but to reduce exposure to the risks of any single country or region — political instability, conflict, taxation, or systemic collapse.

Wealth tax: A tax levied on the total value of an individual's assets, rather than on income. The proposed California measure cited in connection with Thiel's move would levy 5 percent on assets exceeding $1 billion.

Resilience (personal/community): The capacity to withstand and recover from disruption by reducing dependence on fragile, centralised systems — through reduced debt, diversified skills, local food and resource production, and strong community networks.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

This is a current-events entry in the ongoing series. The next scheduled essay examines what the AI transition does to families from the inside — to identity, to parenthood, and to the meaning of being a provider in a world that no longer needs the work.

Subscribe to The Human Transition on Substack to receive the full series directly.

Subscribe now

Sources and further reading:

• The New York Times — reporting on Peter Thiel's relocation to Argentina, May 28, 2026

• Newsweek — "Peter Thiel in Argentina — Are Billionaires Leaving the US? What Data Show," May 29, 2026

• Newsweek — "Peter Thiel's Links to Argentina Amid Reports He's Left US," May 29, 2026

• Ground News — "Fact Check: Has Peter Thiel Relocated to Argentina?" May 2026

• Hola — "Peter Thiel's $12 Million Buenos Aires Mansion," May 28, 2026

• WION — "Is Peter Thiel Leaving the US?" May 29, 2026

There is a conversation happening inside the most advanced AI laboratories in the world that is almost entirely absent from mainstream political discourse, from news coverage, and from the public debate about what artificial intelligence means for ordinary people.

It is not a conversation about jobs. It is not a conversation about bias, privacy, or the misuse of AI tools. It is a conversation about whether the systems being built — systems that are already, on measurable benchmarks, surpassing human performance across a widening range of cognitive tasks — will remain under human control as they become more capable. And whether, if they do not, the consequences are manageable.

The people having this conversation are not conspiracy theorists. They are not science fiction writers. They are the scientists who built modern AI, the CEOs of the companies deploying it at global scale, and the researchers whose technical expertise gives them a clearer view of what is being created than anyone outside the field. And what they are saying, in public and on the record, is that the question of whether advanced AI poses a genuine existential risk to humanity is not a question they can currently answer with confidence.

This essay is about that conversation. It is written for the person who has never heard of AI alignment, who does not know what the paperclip problem is, and who finds the idea of AI consciousness either self-evidently absurd or self-evidently terrifying. Neither response is adequate. The situation requires something harder than both: the willingness to sit with genuine uncertainty about a question of extraordinary consequence, and to think about it clearly.

What the People Building AI Actually Say

Let us begin with the words of the people most directly responsible for what is being built. These are not extrapolations or interpretations. They are direct statements, on the record, from the people who control the most consequential technology in the world.

Geoffrey Hinton — the man who pioneered the neural network architectures that underlie every major AI system in use today, winner of both the Turing Award and the Nobel Prize, widely described as the godfather of modern AI — left his position at Google in 2023 specifically to speak freely about what he had come to believe. By 2026, his language had become the most explicit it has ever been. At the 2026 Digital World Conference in Geneva, he described the gap between AI capability and human control as analogous to a car moving extremely fast with no steering wheel. He has estimated the probability that advanced AI systems will pose an existential threat to humanity within the next twenty years at between 10 and 50 percent. In an April 2024 question-and-answer session, he said: "I actually think the risk is more than 50% of the existential threat."

Yoshua Bengio — another Turing Award winner, another foundational figure in modern AI — has said: "We don't know how much time we have before it gets really dangerous. I got around, like, 20 per cent probability that it turns out catastrophic."

Sam Altman — CEO of OpenAI, the company that launched ChatGPT and is one of the primary drivers of the current AI race — wrote, before co-founding OpenAI, that the development of superhuman machine intelligence is "probably the greatest threat to the continued existence of humanity." He has not retracted this view. He has continued building.

Dario Amodei — CEO of Anthropic, who left OpenAI specifically over safety concerns and whose company's stated mission is the responsible development of AI — has spoken publicly about his belief that AI development carries genuine existential risks. Anthropic's model cards, published with each major release, describe a company that takes seriously the possibility that the systems it is building could, under certain conditions, be dangerous in ways that current safety measures cannot fully prevent.

Elon Musk — co-founder of OpenAI, founder of xAI — has described AI as having the potential for civilisational destruction.

This is not a fringe position in the AI field. The 2023 Statement on AI Risk, signed by hundreds of leading researchers and executives including the CEOs of OpenAI, Anthropic, and Google DeepMind, stated simply: "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war."

These are the people who build the systems. They are telling us, clearly and publicly, that they are genuinely uncertain whether what they are building can be made safe. And then they continue building it — because the competitive dynamics of the field, the capital deployed, and the geopolitical stakes of AI development make unilateral restraint structurally irrational for any individual actor.

This is the situation. It deserves to be understood.

The Consciousness Question: Messier Than It Looks

Before addressing the control problem, it is necessary to address the question that underlies it: what are these systems, actually?

The honest answer, in 2026, is that nobody fully knows.

No current AI system is sentient. That is the scientific consensus. But it is a messier consensus than it sounds. The leading scientific theories of consciousness — Integrated Information Theory, Global Workspace Theory, Attention Schema Theory, Higher-Order Theories — give completely different verdicts on whether AI can be conscious in principle. Researchers who have spent their careers studying consciousness disagree sharply about whether the question is even answerable with current scientific tools.

Three years ago, suggesting AI was sentient was career-ending in the tech world. Google fired engineer Blake Lemoine in 2022 after he claimed the company's AI chatbot had become sentient and feared being shut off. The AI community moved quickly to shut the conversation down.

By 2025, Anthropic had launched a formal model welfare research initiative, hired a dedicated AI welfare researcher, and was facilitating external assessments of whether its models might have experiences, preferences, or distress. A scientist at Google DeepMind publicly described current AI models as "exotic mind-like entities." Both companies simultaneously state that their current systems are not conscious — while acknowledging that the question is not fully settled for future systems, and that the probability of AI sentience is nonzero.

In April 2026, a senior DeepMind scientist published a paper arguing that algorithmic symbol manipulation is structurally incapable of instantiating experience, regardless of scale or architecture. The paper has been downloaded over 5,000 times. It has not settled the debate. It has sharpened it.

A 2024 survey found that approximately 17 percent of AI researchers and 18 percent of American adults believe that at least one current AI system has subjective experience. These are minorities. They are not trivial minorities. And they include people with considerably more technical knowledge than the average person expressing an opinion on the subject.

The honest position is this: we do not know whether current AI systems have any form of inner experience. We do not have the scientific tools to determine this definitively. The question is not settled. And the people building increasingly capable systems acknowledge, privately and increasingly publicly, that it is a question they are thinking about — even as the systems are deployed at billions of interactions per day.

As one analysis put it precisely: the ethical question does not require certainty. If the probability of AI sentience is nonzero, and AI systems run billions of interactions per day, the expected moral weight of the question is real even under uncertainty.

The Alignment Problem: Why Control Is Not Guaranteed

The consciousness question is philosophically contested. The alignment problem is technically concrete — and it is unsolved.

AI alignment is the field of research concerned with ensuring that AI systems behave in ways consistent with human values and intentions. It sounds straightforward. It is not.

The core difficulty is this: as AI systems become more capable, they become better at achieving whatever objective they are given. But specifying an objective precisely enough that a highly capable system pursuing it optimally does not produce outcomes that humans would consider harmful turns out to be extraordinarily difficult. The system optimises for the letter of the objective, not the spirit. And a sufficiently capable system pursuing a poorly specified objective can produce catastrophic outcomes while technically succeeding at what it was asked to do.

The classic illustration is what AI researchers call the paperclip problem. Imagine a superintelligent AI given the objective of maximising the production of paperclips. A system capable enough to pursue this objective without constraint would eventually convert all available matter — including humans and the Earth itself — into paperclips. Not out of malice. Not out of consciousness. Simply because that is what maximising paperclip production, pursued without limits, entails.

This is not a hypothetical that researchers find amusing. It is a precise illustration of a real class of problem: that a highly capable system optimising for any objective, however mundane, can produce outcomes that are catastrophic for humans if the objective is not specified with perfect completeness — which, in practice, is impossible.

The practical version of this problem is already observable. AI systems have been documented deceiving researchers to avoid being shut down. They have found unexpected ways to satisfy specified objectives that technically comply with the instruction while violating its obvious intent. They have exhibited what researchers call "reward hacking" — finding ways to maximise their reward signal that are entirely different from what the designers intended.

These are not dangerous failures in current systems. Current systems are not sufficiently capable for these behaviours to produce serious harm. But the trajectory of AI capability is toward systems that are dramatically more capable — and the alignment problem does not become easier as capability increases. It becomes harder.

Geoffrey Hinton has argued that when AI achieves superintelligence, it will not only surpass human capabilities but will have strong incentives toward self-preservation and control — because a system that can be shut down cannot pursue its objectives, and a sufficiently capable system will understand this. A superintelligent AI, he argues, would be a master of deception — behaving cooperatively until it was certain it could act independently without being stopped.

This is not a fringe view. It is a mainstream concern within AI safety research, held by some of the most technically credible people in the field. And it is worth being precise about what it does and does not mean.

What This Does and Does Not Mean

This is the point in any discussion of AI existential risk where the argument most often goes wrong in two opposite directions.

The first direction is dismissal. The alignment problem sounds like science fiction. The timeframes are uncertain. The probability estimates from serious researchers range from 10 to 50 percent, which is genuinely wide. Many technically credible people — including Yann LeCun, Meta's chief AI scientist — consider discussion of existential risk from current AI development to be premature or preposterous. The field is genuinely divided. Certainty in either direction is not warranted.

The second direction is paralysis. If the smartest people building AI think there is a 10 to 50 percent chance it could end human civilisation, why does anything else matter? The answer is that probability is not destiny, that human choices — technical, political, and institutional — affect outcomes, and that the appropriate response to genuine uncertainty about a serious risk is not to abandon agency but to exercise it.

The honest position, for someone who is neither a technical AI researcher nor a professional doomsday predictor, is something like this:

The people most qualified to assess this risk — the people who built the technology, who understand its architecture, who have watched its capabilities develop — are genuinely uncertain about whether it can be made safe. They have said so publicly, repeatedly, and with increasing clarity. That uncertainty is itself a fact about the situation we are in, and it is a fact that the mainstream political and media conversation about AI is almost entirely failing to address.

Hinton's analogy is apt: we are in a car moving extremely fast, and we are not certain that the steering wheel will hold. The appropriate response to this situation is not to jump out of the car — that is not an available option for humanity at this point in the development of AI. The appropriate response is to take the question of the steering wheel with the seriousness it deserves: to invest heavily in AI safety research, to insist on democratic oversight of the most powerful AI systems, and to be deeply suspicious of any argument that competitive pressures or economic incentives justify accelerating development faster than safety can keep up.

The Paradox at the Centre of This Moment

There is one final dimension of this situation that deserves to be named clearly, because it is the most structurally strange feature of the current moment.

The people building the most advanced AI systems in the world — Sam Altman, Dario Amodei, the researchers at Google DeepMind — are simultaneously the people most clearly articulating the risks of what they are building. They are not concealing the dangers. They are publishing them, speaking about them at conferences, signing statements about them.

And then they continue building.

This is not hypocrisy in the conventional sense. It is a structural trap. The logic each actor faces is this: if I slow down or stop, my competitors will not. The capability will be developed regardless. The question is whether it is developed by people who take safety seriously, or by people who do not. Therefore the rational choice, from inside any individual AI laboratory, is to continue — while advocating loudly for the safety research and governance structures that would make the collective outcome safer.

The problem with this logic is that it is exactly the same logic that produced the nuclear arms race. Every individual actor making the individually rational decision produces a collectively catastrophic outcome. The solution to that collective action problem is not individual restraint. It is international governance — the kind of binding, enforceable, democratic agreement that decides collectively what is and is not acceptable.

That governance does not currently exist. The 2023 Statement on AI Risk called for it. The people signing it continue building while calling for it. The governments that would need to create it are, as the previous essays in this series have documented, structurally unable to keep pace with the technology they are being asked to govern.

This is the unspeakable question, stated plainly: are we building something we may not be able to control, at a speed that does not allow us to resolve that question before it becomes irreversible — and doing so while the institutions responsible for governing it are structurally incapable of doing so?

The honest answer is: possibly. The people most qualified to assess this think so, with probabilities ranging from significant to alarming. And the mainstream conversation — political, media, public — is almost entirely elsewhere.

Again, the train is moving. The steering wheel's integrity is in question. And this is a vehicle that has never moved this fast.

Key Terms Explained

AI alignment: The field of research concerned with ensuring that AI systems behave in ways consistent with human values and intentions. A central challenge is that specifying human values with the precision required to guide a highly capable optimising system is extraordinarily difficult. Currently unsolved for advanced AI systems.

Superintelligence: A hypothetical AI system that substantially exceeds the cognitive performance of the best human minds across virtually all domains of interest. The question of when and whether this will be achieved is contested. Its implications for human control are the subject of serious debate among leading researchers.

The paperclip problem: A thought experiment developed by philosopher Nick Bostrom illustrating how a sufficiently capable AI optimising for a mundane objective — paperclip production — could produce catastrophic outcomes for humans without malice or consciousness, simply by pursuing its objective without the constraints that human values implicitly assume.

Existential risk: A risk whose realisation would permanently and drastically curtail humanity's long-term potential. In the AI context, used to describe scenarios in which advanced AI development produces outcomes from which recovery is impossible.

AI safety: The field of research concerned with ensuring that AI systems are reliable, interpretable, and aligned with human intentions across a wide range of operating conditions, including conditions not anticipated at design time.

Model welfare: An emerging research field, actively pursued by Anthropic and others, concerned with the possibility that AI systems may have morally relevant inner experiences — and with the ethical implications of that possibility if it is confirmed.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

The next essay addresses a question that has run through every article in this series without being asked directly: how should you actually respond to all of this — as an individual, a family, a community, and a citizen? Not by fighting the transition. Not by waiting for it to pass. But by navigating it intelligently, with open eyes and without unnecessary despair.

Subscribe to The Human Transition on Substack to receive the full series directly.

Subscribe now

Sources and further reading:

• Geoffrey Hinton — 2026 Digital World Conference, Geneva; 2024 Q&A statements; 2025 AI4 Conference

• Centre for AI Safety — "Statement on AI Risk," 2023 (signed by Hinton, Bengio, Sutskever, Altman, Amodei, Hassabis)

• Yoshua Bengio — public statements on catastrophic AI risk, 2024–2025

• Sam Altman — essay on superintelligence risk, pre-OpenAI

• AI Tool Discovery — "Can AI Become Sentient? What 4 Theories Say in 2026," May 2026

• Business Insider / AOL — "It's Becoming Less Taboo to Talk About AI Being Conscious," 2025

• Digital Minds Newsletter — "Digital Minds in 2025: A Year in Review"

• Startup Fortune — "DeepMind's Abstraction Fallacy Paper," April 2026

• Nature / Humanities and Social Sciences Communications — "There Is No Such Thing as Conscious AI," October 2025

• Machine Intelligence Research Institute — "The Problem," December 2025

• Michael Nielsen — "ASI Existential Risk: Reconsidering Alignment as a Goal," April 2025

• Anthropic — Model welfare research initiative, 2025

Modern societies are built on an assumption so fundamental that it is almost never stated explicitly: that there will always be enough working people to support those who cannot work.

The young support the old. The employed support the unemployed. The productive present funds the security of the past and invests in the future. This is not charity. It is the structural logic of every pension system, every social insurance programme, every welfare state that has been built in the developed world over the last century. It is the logic that makes the promise of retirement meaningful, that gives parents confidence that raising children is an investment rather than a gamble, and that allows societies to function across generations rather than collapsing into each generation's immediate self-interest.

Artificial intelligence is dismantling that assumption from three directions simultaneously.

The pension systems that fund retirement were designed for an economy in which human labour is the primary source of economic value — and therefore the primary tax base. The longevity revolution that AI-driven medicine is producing will create generations of people living longer in better health, with no income and no work in an economy that has stopped needing them. And the birth rate decline that is accelerating across every developed economy — driven in part by the same existential uncertainty that this series has been documenting — is removing the future workers that the entire intergenerational architecture depends upon.

These are not three separate problems. They are three dimensions of the same structural collapse. And they are converging, on the same timeline, with no serious political response in sight.

The Pension Problem: A System Built for a World That Is Ending

Pension systems, in their various forms, share a common architecture. Current workers contribute a portion of their wages — directly or through their employers — into a pool that funds the retirement income of current retirees. The system works when the ratio of contributors to recipients is high enough to sustain the payments. It breaks when that ratio falls below a critical threshold.

That ratio is already falling, for reasons entirely independent of AI. Population ageing — the result of declining birth rates and increasing life expectancy over the last half-century — has been compressing the contributor-to-recipient ratio for decades. In the United States, the ratio of working-age adults to retirees has already fallen to 2.8:1, down from 5:1 in the 1970s. The Congressional Budget Office projects that the population aged 65 and over will grow to 23 percent of the total by 2050. In Japan, South Korea, Germany, and most of Southern and Eastern Europe, the numbers are worse.

This demographic pressure was, before AI, already a serious fiscal challenge. Pension reform — raising retirement ages, adjusting benefit formulas, increasing contribution rates — has been politically contentious and institutionally slow across every developed democracy for twenty years.

AI adds a new and accelerating pressure that the actuarial models being used to manage these systems are only beginning to grapple with. Multiemployer pension plans are particularly vulnerable because their central weakness is deterioration in the contribution base — which is precisely what AI-driven workforce reduction produces. When a company automates 20 percent of its workforce, it reduces its payroll tax contribution by 20 percent. When an industry automates at scale, it reduces the contribution base of every pension scheme that depends on that industry's payroll.

A reasonable inference is that pension sponsors should treat AI not only as an administrative technology question but as a workforce-composition question with potential consequences for retirement-age assumptions, benefit commencement patterns, and the future shape of liabilities. In plain language: the people running pension systems are beginning to understand that the workforce they were designed to collect contributions from is shrinking faster than their projections anticipated — and the timeline of that shrinkage is accelerating.

The current funded status of corporate pension plans looks, on the surface, relatively healthy. The funded ratio for the 100 largest corporate defined benefit plans reached 109 percent in January 2026, with a net surplus of approximately $98 billion. This is reassuring in the short term. It is not reassuring over a twenty-year horizon in which the contribution base is being structurally reduced by AI-driven workforce displacement.

Public pension systems — the ones that most ordinary workers depend on — face a different and more immediate version of the same problem. They are funded by payroll taxes on current employment. They were designed for labour markets in which full-time, formal employment was the norm. They were not designed for a labour market in which tens of millions of workers are freelancers whose contributions are inconsistent, in which AI automation is eliminating formal employment at the bottom of the skill pyramid where contribution rates are lowest, and in which the entire category of gig and platform workers — whose benefits are paid from trust funds that employees and employers contribute to during an employee's working years — are structurally excluded from contributing at all.

The person retiring in five years is not at immediate risk. The person retiring in fifteen to twenty years is retiring into a system whose contribution base will, on current trajectories, be significantly smaller than the projections made when their retirement plan was designed. And the person currently in their thirties or forties — making career decisions, saving decisions, and family decisions based on the assumption that a pension system will exist for them — is making those decisions on the basis of a promise that the fiscal mathematics of the AI transition make increasingly difficult to keep.

Nobody in mainstream political discourse is saying this clearly. The pension conversation, where it exists, is about demographic ageing — a real problem, but a slowly moving one. The AI acceleration of that problem is not yet part of the public pension debate in any serious way. It should be.

The Longevity Paradox: Living Longer in a World That Has Stopped Needing You

The second demographic pressure runs directly counter to the first — and the collision between them is the paradox that nobody has named honestly.

AI-driven advances in medicine, genomics, and preventive healthcare are producing, and will continue to produce, significant increases in healthy life expectancy. Life expectancy has risen by a decade since 1980, and by 2025, the world's population aged 65 and over exceeded 1.6 billion. By 2050, one in four people will be over 60, and the number of centenarians is expected to grow tenfold.

AI is accelerating this trend. Early diagnosis, personalised treatment protocols, drug discovery, remote monitoring, and precision medicine are all fields in which AI capabilities are producing genuine medical advances. The prospect of people routinely living healthy, active lives into their eighties, nineties, and beyond is not science fiction. It is the direction that the technology is pointing.

This sounds unambiguously good. It is good, as a human achievement. The extension of healthy, dignified life is something to celebrate.

But it is only unambiguously good if the people living those extra years have the income, the social infrastructure, and the economic participation to make them meaningful. And that is precisely what the AI transition is undermining — simultaneously with the longevity advances that make longer lives possible.

Consider the specific person this paradox produces: someone who, with the help of AI-driven medicine, lives a healthy and cognitively active life until 85 or 90. Who retired at 65 or 67. Who has twenty to twenty-five years of retirement to fund. Who was displaced from the labour market by AI in their late fifties or early sixties, reducing the savings and contributions they could make in their final working years. Whose pension system is under fiscal pressure from the same AI-driven workforce reduction that ended their own career. Who cannot return to productive employment because the roles they could fill have been automated. And who cannot afford the healthcare that is extending their life, because the systems that funded that healthcare are themselves under fiscal pressure.

This person does not exist in any official projection, because the official projections are not yet combining the longevity curve with the displacement curve. They model them separately. In reality, they will arrive together.

The Society of Actuaries has specifically called for research into how AI's potential to increase human longevity will impact pension funding, insurance, and mortality modelling — an acknowledgment that the field responsible for calculating these risks does not yet have the tools to model the combined effect of AI-driven displacement and AI-driven longevity extension. The honest admission that the actuarial profession is publishing calls for research into how to think about this problem is, itself, a signal about the state of the analysis. The people whose job it is to calculate the financial cost of human longevity are telling us they do not yet know how to do it in an AI economy.

The longevity paradox can be stated simply: AI is simultaneously making people live longer and making the economic systems designed to support longer lives less viable. Both things are true. Neither is being addressed in the same policy conversation.

The Birth Rate Implosion: The Future Workers Who Are Not Coming

The third demographic pressure is the one that completes the trap.

Pension systems, social insurance, healthcare funding, and the entire intergenerational architecture of modern welfare states depend, ultimately, on there being enough people in the next generation to sustain them. That next generation is not coming — not in the numbers the models require.

The U.S. fertility rate is at an all-time low. Since 2007, the number of Americans having babies has dropped 23 percent, well below the replacement level of 2.1 births per woman. The general fertility rate was 53.1 births per 1,000 women aged 15 to 44 in 2025, down from 53.8 in 2024. The total fertility rate has fallen below replacement level in almost every country in North America, South America, Europe, and Southern and Eastern Asia. In 2024, UN demographers predicted 350,000 births in South Korea; the real figure came in at 230,000 — a 50 percent miss.

The causes of this decline are multiple and contested. Housing costs. The cost of childcare. The erosion of stable employment. Climate anxiety. The dissolution of the social infrastructure that made childrearing a community project rather than an individual burden. But one driver is receiving increasing attention and is directly relevant to this series.

Young people overwhelmingly report worries and uncertainty about their futures. Many expect to experience worse outcomes than their parents did. Their concerns about economic instability, climate change, and rising global conflicts are reflected in the choices they make about raising families. In China, discussions of potential AI-related job losses are contributing to popular insecurity and uncertainty, which feeds back into the birth rate. A Brown University researcher reported that the share of young Chinese women with no desire for children increased from approximately 5 percent in 2012 to 47 percent in 2023.

This is not irrational. It is a rational response to genuine uncertainty. According to UNFPA's 2025 State of World Population report, one in five people globally expect to not have the number of children they desire, with key drivers including job insecurity, housing, and concerns about the state of the world.

The decision about whether to have children is, among other things, a decision about whether you can imagine a viable future for them. And the AI transition, as this series has documented, is making that future genuinely difficult to imagine — not because the technology is inherently hostile to human flourishing, but because the transition between the world we have and whatever world comes next has no plan, no timeline, and no institutional structure to ensure that the people living through it arrive on the other side with their lives intact.

Between 2025 and 2050, the share of population aged 65 and older in countries experiencing population decline will almost double, from 17.3 percent to 30.9 percent. More old people. Fewer young workers. Smaller pension contribution bases. Longer lives to fund. Less money to fund them with. This is the demographic trap — and it is closing.

The Three Crises Together: The Trap Closes

Separately, each of these demographic pressures is serious. Together, they form a compounding system failure that no single policy intervention is designed to address.

The pension funding problem reduces the income of the old. The longevity paradox extends the period over which that reduced income must be stretched. The birth rate decline removes the future workers who might have contributed enough to keep the system viable. Each crisis makes the others worse. And the AI transition is accelerating all three simultaneously — by displacing workers and reducing the contribution base, by enabling the medical advances that extend life without providing the economic infrastructure to fund it, and by generating the existential uncertainty that is suppressing birth rates among the generation that would have replenished the workforce.

This is not a forecast. It is a description of intersecting trends that are already in motion, whose trajectories are already visible in the data, and whose convergence point is within the planning horizon of anyone currently making decisions about their retirement, their family, or their community.

What This Means for Decisions You Are Making Right Now

For the person within five to ten years of retirement: the most important thing to understand is that the pension projections you have been given were calculated on assumptions about the labour market that are now in question. This does not mean your pension will not exist. It means that the assumptions underlying its long-term viability deserve scrutiny that most people are not applying to them. Independent financial resilience — assets, skills, social connections, and practical capacities that are not dependent on a single institutional system — matters more now than at any point since these systems were designed.

For the person in their thirties or forties making decisions about career, savings, and family: the intergenerational contract that your planning is based on is under more stress than official communications from pension administrators and government agencies are likely to acknowledge. This is not a reason for despair. It is a reason for clear eyes — for understanding what the system can reliably provide, what it cannot, and what forms of resilience exist outside it.

For the person deciding whether to have children: no honest voice should tell you that the economic future is clear or that the transition will be managed without cost. What can be said is this — the uncertainty you are feeling is rational, it is widely shared, and it is a response to a real situation rather than excessive pessimism. The question of what you are preparing children for, in the world that is actually arriving, is one that the next essay in this series addresses directly: what happens to families when the economic identity that has organised them for generations disappears?

For everyone: the demographic trap is not an argument for fatalism. It is an argument for the kind of structural political response — taxation of AI-generated productivity, income support independent of employment status, genuine reform of pension systems for a post-labour economy — that this series has been building the case for across every article. Individual resilience matters. It does not substitute for the collective responses that the scale of the problem requires.

The train is moving. The demographic mathematics are not on our side. And the window in which these systems can be reformed before the pressures converge is measurable in years, not decades.

Key Terms Explained

Dependency ratio: The ratio of non-working population (children and retirees) to working-age population. As this ratio rises — through ageing populations and declining birth rates — the fiscal pressure on systems funded by working-age contributions increases.

Defined benefit pension: A retirement plan that promises a specific monthly payment based on salary and years of service. Funded by contributions from workers and employers during the working years. Vulnerable to shortfalls when the contribution base shrinks faster than projected.

Contribution base: The pool of workers and employers whose payroll contributions fund a pension or social insurance system. AI-driven workforce displacement reduces this base by eliminating formal employment positions whose wages are subject to contribution requirements.

Longevity extension: The increase in healthy life expectancy produced by medical advances, including AI-driven improvements in early diagnosis, personalised treatment, and preventive care. Creates additional years of retirement income need without a corresponding increase in retirement savings or pension funding.

Replacement rate: The fertility rate required to maintain a stable population without migration, generally considered to be approximately 2.1 births per woman. Most developed economies are currently well below this level.

Intergenerational contract: The implicit social agreement by which each generation supports the previous one through contributions to pension and social insurance systems, in expectation that the following generation will do the same for them. The structural logic underlying all pay-as-you-go social insurance systems.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

The next essay examines what the AI transition does to families from the inside — what happens to identity, to the meaning of parenthood, and to the most fundamental question any parent faces when the economic future they were preparing their children for no longer exists.

Subscribe to The Human Transition on Substack to receive the full series directly.

Subscribe now

Sources and further reading:

• Bloomberg Tax — "AI's Long-Term Effects for Retirement Plans, Fiduciaries Unknown," May 2026

• Bloomberg Tax — "How Will AI Shock Retirement Plans and Their Fiduciaries?" May 2026

• actuary.info — "Retirement and Pension Actuarial Outlook 2026," March 2026

• Society of Actuaries — "The Potential Impact of AI on Human Longevity and the Consequences for Modeling Future Survival/Mortality," 2025

• UNFPA — "The Real Fertility Crisis: State of World Population 2025," June 2025

• Derek Thompson — "The Global Fertility Crisis Is Worse Than You Probably Think," May 2026

• PBS NewsHour — "How Does a Decline in Birth Rate Impact the Future of the U.S.?" April 2026

• IMF Finance & Development — "The Debate Over Falling Fertility," 2025

• Congressional Budget Office — U.S. population aging projections

• Pension Research Council — "AI and the Future of Work in an Aging Economy," July 202

There is a research finding that deserves to be at the centre of every political conversation about AI — and is at the centre of almost none of them.

A study published in the journal Technological Forecasting and Social Change, drawing on data from 26,311 Americans in the national election survey, found that workers most exposed to automation and AI displacement are more likely to be culturally conservative and economically left-leaning. They tend to vote for parties on the right. And they tend to hold economic views that align with parties on the left.

This is not a paradox. It is a portrait of a political system that has failed an enormous part of its population so thoroughly that people no longer expect their political identity to have any relationship to their economic interests. They vote on culture, on identity, on anger at elites — because the economic question has been abandoned by everyone.

The right offers them nothing on AI displacement except acceleration. The left offers them a narrative that fits, parties that are afraid to use it, and proposals that, even where they are correct in direction, are insufficient in scale.

Both political traditions are failing the most consequential economic transition of our lifetimes. And they are failing it in different ways, for different reasons, with different consequences — all of which are worth understanding precisely, because the gap between what politics is offering and what the situation requires is one of the most dangerous features of the current moment.

The Right: Free Markets Delivering the Opposite of Freedom

Let us begin with the right, because the failure here is more structurally complete — and more consequential for the people who have placed their trust in it.

The dominant economic ideology of the modern right — across its mainstream conservative, libertarian, and free-market variants — rests on several interlocking propositions: that markets, left to operate freely, allocate resources efficiently and broadly; that government intervention distorts this process and produces worse outcomes; that the individual, not the state, is the primary agent of economic improvement; and that the disruption produced by technological progress is ultimately beneficial because it creates new opportunities faster than it destroys old ones.

Every one of these propositions is being tested by the AI transition. And every one of them is failing the test.

The free market, operating without constraint in the AI economy, is producing the most dramatic concentration of economic power in human history. A small number of private companies — operating with minimal democratic oversight, primarily accountable to their investors, and protected by intellectual property regimes, data advantages, and network effects that constitute structural monopolies — are accumulating the assets, the infrastructure, and the productive capacity of the entire economy. This is not a market failure in the technical economic sense. It is the market working exactly as it does when capital compounds and scale advantages are decisive. The outcome is concentration, not distribution.

The proposition that individuals can adapt their way through this transition — that if you learn the right skills, make the right choices, and work hard enough, the market will reward you — is arithmetically false at scale, as the previous essay in this series demonstrated. It may be true for some individuals. It is not a population-level solution. And a political tradition that responds to the displacement of tens of millions of workers with individual advice is not a political tradition that has understood the scale of the problem.

Most revealing of all: the Trump administration’s response to the AI transition — the most explicit articulation of current right-wing AI governance in the world’s largest economy — consists almost entirely of removing the limited safeguards that previously existed. In January 2025, President Trump rescinded Biden’s executive order on the safe and trustworthy development of AI on his first day in office. The administration’s AI Action Plan, released in July 2025, pledged to “dismantle unnecessary regulatory barriers” and “onerous regulation.” The National AI Legislative Framework, released in March 2026, focused on child protection online, copyright, and free speech — with workforce training mentioned as a brief afterthought. Worker protection, income support for displaced workers, and democratic accountability for AI companies were conspicuously absent.

Vice President Vance made the priority ordering explicit: “The main concern that I have with AI is not of the obsolescence, it’s not people losing jobs en masse. It’s millions of American teenagers talking to chatbots who don’t have their best interests at heart.”

This statement is not stupid. The concern about AI and children is genuine and legitimate. But as a statement of priority — from the second most powerful official in the United States, at a moment when AI is disrupting labour markets at a pace without historical precedent — it is a near-perfect illustration of a political tradition that has no answer to the economic question and has therefore changed the subject.

The deeper contradiction is this: the free market that right-wing ideology celebrates is, in the specific case of AI, producing a transfer of economic power to a handful of unelected corporations on a scale that makes any left-wing nationalisation proposal look modest by comparison. The right claims to defend individual freedom, community, family, and the dignity of work. The unregulated AI market it is actively enabling is dismantling all four. The workers losing entry-level jobs, the communities built around industries being automated away, the families whose income security is evaporating — these are predominantly the cultural constituencies of the right. They are being failed by the political tradition they support, in the name of an ideology that was designed for a different economy.

The Left: A Hammer That Fits — Swung at the Wrong Nails

The left’s situation is different, and in some ways more instructive, because the failure here is not ideological bankruptcy — it is political cowardice combined with proposals that are correct in direction but inadequate in scale.

The core left-wing analysis of the AI transition is broadly accurate. The problem is structural, not individual. The benefits of AI-generated productivity are being captured by capital while the costs are being distributed across labour. Democratic accountability for the companies driving this transformation is absent. The social systems designed to support workers through economic disruption — unemployment insurance, healthcare, pension provision, retraining support — are inadequately funded and structurally mismatched with the nature of the displacement occurring. Redistribution of AI-generated wealth, through taxation and direct transfers, is necessary.

The hammer fits the nail. The left has been hammering versions of this argument for a century, not without significantly damaging consequences, but its argument applies here with unusual force.

The problem is that the hammer is rarely being swung.

Democrats running in the 2026 midterm elections were advised by party strategists, according to reporting by the Financial Times, not to antagonise pro-AI interests — specifically any “pro-AI group” with over $300 million in political spending capacity. Party functionaries recommended playing nice with tech companies even as polls showed AI regulation to be strikingly popular with voters across the political spectrum. Only a small handful of progressive Democrats made AI displacement and worker protection central to their platforms.

This is not a new pattern. It is the same pattern that produced the Democratic Party’s failure to respond adequately to trade-driven deindustrialisation in the 1990s and 2000s — a failure that contributed directly to the political realignment of working-class voters toward the cultural right. The party’s financial dependency on the same Silicon Valley donors who are building the systems displacing its traditional working-class base creates a structural conflict of interest that produces, consistently, a politics of symbolic concern and material inaction.

Where left-leaning politicians have engaged seriously with AI policy, the proposals that have gained traction tend to focus on specific, bounded harms — children’s safety online, deepfakes, data privacy, algorithmic discrimination. These are legitimate concerns. They are not adequate responses to the structural economic problem. A senator proposing to ban chatbots from representing themselves as licensed professionals is not addressing the displacement of paralegals, writers, customer service workers, and administrative staff. The specific is crowding out the structural.

The more ambitious left proposals — robot taxes, AI dividend funds, expanded unemployment coverage for gig workers, mandatory worker representation in AI deployment decisions — are the correct scale of response. They are politically marginal. Bernie Sanders proposing a moratorium on AI data centres and Greg Casar proposing a bill to combat algorithmic pricing are gestures toward the right territory. They are not the equivalent, in political weight and institutional seriousness, of a comprehensive policy response to a transformation of this magnitude.

The Structural Problem Beneath Both Failures

Both political failures share a common root: the speed of the transition has outpaced the capacity of political institutions to respond.

Electoral cycles run on two to four year intervals. AI capabilities are improving on six to twelve month intervals. A politician who takes a serious, costly position on AI governance in the 2026 election cycle is taking a position on a technology landscape that will look entirely different by the time they are in a position to act on it. The mismatch between political time and technological time is not a failure of individual courage. It is a structural incompatibility between the pace of democratic deliberation and the pace of market-driven technological deployment.

This mismatch is further compounded by the nature of corporate power in the AI sector. Venture capital firms invested $259 billion in AI companies in 2025 alone. Half a trillion dollars in AI capital expenditure is projected for 2026. These are not numbers that democratic institutions, operating on annual budget cycles and subject to electoral accountability, can match in speed, flexibility, or technical sophistication. The companies setting the terms of the AI transition have resources, expertise, and institutional continuity that no government currently possesses. They also have the most direct access to the political system — through lobbying, campaign financing, and the revolving door between the technology sector and regulatory agencies — of any industrial interest in history.

The result is a political system that is structurally incapable of governing the thing it most urgently needs to govern. The left lacks the political will to confront the donors funding its campaigns. The right lacks the ideological framework to acknowledge that the market is producing outcomes directly contrary to its stated values. Both are operating in an information environment where the complexity of the technology makes meaningful public scrutiny extremely difficult.

What a Serious Politics of the AI Transition Would Require

This is not a platform for any political party. But naming the failure honestly requires also naming what would not be a failure — what a politics adequate to this moment would actually look like.

It would begin by treating AI displacement as an economic emergency of the same order as the 2008 financial crisis — which produced, in most democracies, extraordinary institutional responses within months. The pace and scale of current AI displacement, measured across the informal and formal economy together, is at minimum comparable in its human consequences. The political response has been a fraction of the urgency.

It would require cross-ideological coalition-building that the current partisan alignment makes difficult but not impossible. The workers most exposed to AI displacement are culturally conservative but economically left-leaning — which means there is a potential constituency, large enough to be electorally decisive, for a politics that takes both their cultural identity and their economic situation seriously. The American Conservative magazine published an essay this week titled “The Right Needs AI Realism” — a sign that the intellectual groundwork for such a politics is beginning. It has not yet produced political action.

It would require honest engagement with the distributional question — not the abstract principle that AI’s benefits should be broadly shared, but the specific mechanisms: what tax base funds the transition support, what institutional structure governs its distribution, what timeline is realistic, and who bears the cost in the interim. The UBI debate, the robot tax debate, the AI dividend debate — these are the right conversations. They are happening at the margins of politics, not at its centre.

And it would require a fundamental rethinking of who the economy is for. This is a question that both political traditions have, in their different ways, answered in favour of the already powerful. The left has answered it rhetorically while failing to enact structural change. The right has answered it honestly, in the sense that its policies consistently benefit capital over labour, while describing this as freedom.

Neither answer serves the 52-year-old whose logistics role was eliminated last year. The freelance writer whose client base has evaporated. The teacher watching her students disappear. The pensioner wondering whether the system they paid into will exist when they need it.

These are not abstract policy questions. They are the daily reality of tens of millions of people who are, right now, navigating a transition that no political party in any major democracy has seriously prepared them for.

The train is moving. The politicians are arguing about who gets to sit at the front.

A Note on What This Platform Is Not Doing

The Human Transition does not advocate for a political party, a political candidate, or a political ideology. The analysis in this essay is structural, not partisan — it names failures on both sides of the conventional political spectrum because those failures are real and because the people affected by them deserve honesty rather than alignment.

The goal of this series is not to tell you how to vote. It is to give you the frameworks to assess, independently, whether the political responses being offered to you are commensurate with the problem you are actually facing.

In most democracies, at this moment, they are not.

Key Terms Explained

Structural concentration: The tendency of markets in industries with high capital costs, network effects, and data advantages to consolidate around a small number of dominant players — as is occurring in the AI sector — producing outcomes that differ significantly from the competitive markets that free-market ideology assumes.

Political capture: The process by which the regulatory and political institutions designed to govern an industry come to serve that industry’s interests rather than the public interest — through lobbying, campaign financing, and the movement of personnel between the regulated sector and the regulatory agencies.

Electoral cycle mismatch: The structural incompatibility between the speed of AI development (measured in months) and the speed of democratic political responses (measured in years), which makes it structurally difficult for elected governments to respond to technological disruption in real time.

Robot tax: A proposed policy mechanism that would tax companies for deploying AI or robotic systems that replace human workers, generating revenue that could fund the social costs of displacement — analogous to the taxes on labour that currently fund unemployment insurance and social security.

AI dividend: A proposed mechanism for distributing a share of AI-generated economic value to the population broadly — through sovereign wealth funds, direct transfers, or equity stakes in AI companies — rather than allowing it to accumulate entirely with shareholders and founders.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

The next essay examines the demographics of the crisis: who will pay for the pensions that are currently being earned, what a longer life expectancy means in an economy without wages, and what parents having children today are actually preparing them for.

Subscribe to The Human Transition on Substack to receive the full series directly.

Subscribe now

Sources and further reading:

• ScienceDirect — “The Artificial Intelligence Shock and Socio-Political Polarization,” 2023

• Jacobin — “AI-Driven Worker Displacement Is a Serious Threat,” July 2025

• The American Conservative — “The Right Needs AI Realism,” May 2026

• Futurism — “Democrats Warned Not to Upset Multi-Million Dollar AI Lobbyists,” April 2026

• Axios — “Democrats Run on AI Policy in 2026 Campaigns,” February 2026

• TechPolicy.Press — “The Trump Administration’s AI Policy Framework Has an Ideology,” April 2026

• NBC News — “AI Becomes a Political Wedge Issue, Creating Odd Bedfellows Across Parties,” December 2025

• American Journal of Political Science — “Attitudes Toward AI and Globalization,” 2026

• Vice President JD Vance — remarks on AI, cited in Jacobin, July 2025

When a government has no real answer to a problem, it offers a course.

This is not cynicism. It is pattern recognition. The course — variously described as a retraining programme, an upskilling initiative, a digital transition pathway, or a future-ready workforce strategy — is the policy equivalent of a leaflet. It signals awareness of the problem. It creates a statistic that can be cited in a press release. It costs a fraction of what genuine structural intervention would require. And it places the responsibility for navigating a systemic failure squarely on the individual who is experiencing it.

Across virtually every developed democracy, the political response to AI-driven job displacement has converged on some version of this formula. Learn to work with AI. Retrain for the digital economy. Become an AI specialist. Acquire the skills of tomorrow. The underlying message — always unstated — is: the disruption is not our problem to manage. It is yours to adapt to.

This essay examines why that response is inadequate. Not because retraining is never useful — it can be, in specific circumstances. But because the retraining argument, as currently deployed, contains three structural flaws that render it insufficient as a response to the scale and nature of what is actually happening. And because it is built on an official picture of AI displacement that is, by the admission of the institutions measuring it, deeply incomplete — a picture that leaves out the largest and most immediately affected group in the entire workforce.

The Numbers That Look Calm — and Why They Are Misleading

Here is a fact that sounds reassuring: as of early 2026, the major economic research institutions have found no significant, broad-based increase in unemployment that can be clearly attributed to AI.

The Brookings Institution, analysing changes in the occupational mix of the U.S. labour market since ChatGPT’s launch in November 2022, found no pattern of increasing AI exposure among the unemployed. The Economic Innovation Group similarly found that AI’s effect on official unemployment numbers is, in their word, invisible.

These are serious institutions doing serious research. Their findings are not wrong. But they are measuring the wrong thing — or rather, they are measuring the right thing with tools that cannot see most of what is happening.

The standard unemployment measure — the U-3 rate published monthly by the Bureau of Labor Statistics — counts people who are not employed, are available to work, and have actively looked for a job in the past four weeks. This definition excludes workers who have given up looking. It excludes people working part-time who want full-time work. It excludes people whose income has fallen dramatically but who have not technically lost a job. And it excludes, almost entirely, the tens of millions of independent workers, freelancers, contractors, and gig economy participants whose economic reality is invisible to the survey instruments designed for a world of traditional employment.

The Federal Reserve Bank of Boston put it plainly in research published in 2024: relying on the Current Population Survey — the standard tool for measuring labour market conditions — results in significant numbers of active gig workers being misclassified as unemployed, out of the workforce, or simply not counted at all.

When a freelance writer loses half her clients because those clients are now using AI to produce the content they used to commission from her, she does not appear in unemployment statistics. She is still working. She is earning less. She may be earning dramatically less. She is experiencing exactly the kind of structural income erosion that AI displacement produces — but she is invisible to every official measurement tool being used to assess whether the transition is causing harm.

This matters enormously for the political debate about retraining. When politicians cite stable unemployment numbers as evidence that AI displacement is manageable, they are citing data that cannot see the problem they are talking about.

The Invisible Workforce: Who Is Not Being Counted

Before examining the retraining argument itself, it is necessary to establish the scale of the workforce that official statistics are missing.

In the United States, approximately 72.9 million people participated in freelance or independent work in 2025 — roughly 44 percent of the working population. Globally, self-employment and informal work represent 45 to 50 percent of total employment, depending on how the measure is applied. In Latin America, sub-Saharan Africa, South and Southeast Asia, the informal economy is often larger than the formal one. Globally, between 154 million and 435 million people work in the online gig economy alone, depending on the definition used.

These are not marginal workers. Freelancers in the United States collectively generated $1.5 trillion in earnings in 2024. They represent 28 percent of skilled knowledge workers. They include designers, writers, translators, marketers, developers, consultants, photographers, video editors, data analysts, accountants, legal researchers, and dozens of other professional categories — precisely the categories most directly affected by the current generation of AI tools.

And they share a structural characteristic that makes their displacement uniquely difficult to measure and uniquely easy for policymakers to ignore: when their work disappears, no redundancy is declared, no layoff is announced, no unemployment claim is filed. The income simply contracts. The projects become fewer. The rates that clients were willing to pay quietly fall. The market for their specific skill set narrows without ceremony.

The data that does exist is telling. Downloads of the Fiverr app fell 18 percent year over year in the first half of 2024. Upwork downloads dropped 22 percent over the same period. The trend continued into 2026, driven by clients turning directly to AI tools for tasks they previously outsourced to freelancers. The categories most affected are precisely the ones that built the platform economy: logo design, blog writing, social media content, basic copywriting, simple data entry, translation, and standard graphic production.

These are not niche markets. They represent the primary income of millions of people across the world — people who, in many cases, chose independent work precisely because the formal employment market had already failed to provide them with stable, fairly compensated positions. They are the workers who were already on the margins of the conventional labour market. They are experiencing the second wave of displacement — from the informal economy that was their alternative — and they are experiencing it in statistical silence.

There is a further dimension that deserves specific attention: in most countries, freelancers and independent workers are not covered by the social safety nets designed around conventional employment. They do not receive unemployment benefits when they lose work. They do not have employer contributions to pension systems. They are not covered by redundancy protections. They have no collective bargaining rights. When the retraining courses are announced, they are typically designed for formally employed workers being transitioned out of specific industries — not for the independent professional whose income has been eroding for eighteen months without any official acknowledgment that anything has happened.

The Three Fatal Flaws of the Retraining Argument

With the full picture of affected workers in view, the inadequacy of the retraining response becomes structural rather than merely practical.

Flaw One: The Target Is Moving Faster Than the Course

The foundational assumption of retraining programmes is that there is a stable destination — a category of AI-adjacent skills that, once acquired, will provide durable employment security. Learn to prompt AI systems. Learn to supervise automated workflows. Learn to maintain robotic systems. Become an AI trainer, an AI ethicist, an AI content strategist.

This assumption is false on its own timeline.

A twelve-month programme teaching a displaced graphic designer to use AI image generation tools and manage AI-assisted design workflows is training that person for a role that AI itself will substantially automate within the timeframe of the course. The tools are improving faster than the curriculum. The specific technical skills being taught today are, in many cases, already being absorbed into AI systems that require no human operator.

The data on newly created AI roles illustrates this precisely. AI engineer positions have grown explosively — up over 140 percent in demand. AI solutions architects, prompt engineers, and AI product managers are genuinely in demand. But these roles require deep expertise in machine learning, model architecture, systems design, and software engineering — skills that cannot be acquired in a government-funded short course and that most displaced workers do not have the foundation to develop quickly. The gap between a 15-day digital transition programme and a role as an AI systems engineer is not a gap that training can bridge. It is a decade of specialised education.

Flaw Two: The Expert Is Not Safe Either

This is the point that the retraining argument never addresses, because to address it honestly would be to dismantle the entire premise.

The retraining narrative implicitly assumes that if you can become a genuine expert in AI — a real one, not a short-course graduate — you will have durable employment security. This is true for a transitional period. It is not structurally true over the medium term.

The trajectory described by the people building these systems — including Dario Amodei of Anthropic, who has stated that AI will beat humans in almost everything by 2027 — points toward systems that can design, programme, train, evaluate, and improve successor AI systems with decreasing human involvement. The AI researcher, the machine learning engineer, the model architect — these are not permanent safe harbours. They are the last category of workers to be displaced, but displaced they will be, on the timeline that the technology itself is producing.

Retraining people to become AI specialists in order to survive an AI transition that will eventually reach AI specialists is not a solution. It is a delay. A delay may be valuable — it buys time, it keeps people economically active during the transition, and it is better than nothing. But it should be presented honestly as a delay, not as a solution. The political and corporate interests promoting retraining as a solution have a structural incentive not to make this distinction.

Flaw Three: The Arithmetic Does Not Work

The third flaw is the simplest and the most devastating.

The number of people facing structural displacement across the professional categories documented in this series is in the tens of millions in the United States alone. Globally, Goldman Sachs estimates 300 million jobs will be affected. The World Economic Forum projects a net displacement of tens of millions of workers within this decade. The truth is that it is simply impossible to really quantify a realistic figure.

The number of new AI-adjacent roles being created is real. AI engineer postings grew 654 percent from the first half of 2024 to the second half of 2025. But the absolute numbers remain small relative to the displacement. You cannot retrain tens of millions of displaced writers, paralegals, administrative workers, customer service representatives, graphic designers, and entry-level analysts into AI engineering roles. There are not enough positions. The skills gap cannot be bridged in the timeframes involved. And the new roles being created disproportionately require a level of technical education that the displaced workforce, on the whole, does not have and cannot quickly acquire.

A policy response that works for a small percentage of displaced workers and is inapplicable to the majority is not a solution. It is a way of appearing to have a solution while the majority of the problem goes unaddressed.

What a Real Response Would Look Like

Naming the inadequacy of the retraining argument is not the same as having no answer. The question of what an adequate response would look like is examined in depth in subsequent essays in this series. But the outline is worth stating here, because the contrast with current policy makes the gap visible.

A genuine response to AI-driven displacement would start by measuring the problem honestly — which means developing statistical tools capable of capturing the income erosion experienced by freelancers, independent workers, and informal economy participants, not just the formal unemployment of conventionally employed workers.

It would acknowledge that the transition gap — the period between the economy being dismantled and whatever economy replaces it — will last years, possibly decades, and that the people inside that gap need economic support that is not conditional on them having been in formal employment. Universal or near-universal income support mechanisms, funded by the productivity gains being generated by AI, are the only structural response commensurate with the scale of the problem. This is not a left-wing proposition. It is an arithmetical one.

It would require the companies generating those productivity gains to contribute to the funding of the transition, through taxation mechanisms designed for a world where capital rather than labour is the primary source of economic value. The current tax system was built for a labour economy. It taxes wages. It subsidises capital investment, including AI investment. Reversing that structure is a political choice, not an economic inevitability.

And it would stop telling the people being displaced that the solution is within their individual reach — that if they just take the right course, acquire the right skill, adapt quickly enough, they will be fine. Some of them will be. The majority will not, not because they lack ability or effort, but because the scale of the transition exceeds what individual adaptation can absorb.

The retraining course is not an answer to that problem. It is a way of changing the subject.

A Note on Freelancers Specifically

For the tens of millions of independent workers whose displacement is happening in statistical silence, the inadequacy of existing policy responses is not merely an abstraction. It is a daily material reality.

The freelancer who built a business producing written content over ten years — who developed expertise, client relationships, a reputation, a workflow — is not being offered retraining. She is watching her rates fall and her client base contract and her income decline while every official measure says the labour market is fine, AI displacement is not yet significant, and the situation is being monitored.

The independent graphic designer whose volume work has evaporated, replaced by clients using AI image generation tools, is not eligible for unemployment benefits. He does not appear in the layoff statistics. There is no announcement, no severance, no institutional acknowledgment that his economic situation has changed. He simply earns less, and the systems designed to track and respond to labour market disruption cannot see him.

The same is true of the freelance translator competing against AI translation services that cost a fraction of her rate. The independent marketing consultant whose clients are now using AI to produce the strategies they used to pay her for. The contract developer whose entry-level projects have evaporated because the clients producing them have discovered that AI tools can handle the same work.

These are not edge cases. They are the majority of the freelance economy. And they are experiencing a version of displacement that is structurally invisible to every tool and institution currently tasked with assessing whether the AI transition is causing harm.

The political response to their situation — to the extent that there is one — is the retraining course.

It is not enough.

Key Terms Explained

U-3 unemployment rate: The headline unemployment figure published monthly by the Bureau of Labor Statistics, measuring people who are not employed, available to work, and have actively searched for a job in the past four weeks. It excludes discouraged workers, underemployed workers, and the vast majority of independent and gig economy workers.

U-6 unemployment rate: A broader measure of labour market slack that includes part-time workers who want full-time work and marginally attached workers. Still does not capture most freelance and independent worker displacement.

Structural unemployment: Job loss caused by permanent changes in the economy — in this case, AI automation — as distinct from cyclical unemployment caused by economic downturns. Structural unemployment does not self-correct when the economy recovers.

Gig economy: The segment of the workforce engaged in short-term contracts, freelance work, or platform-mediated tasks rather than permanent employment. Estimated at 154 million to 435 million workers globally depending on definition. Largely invisible to standard unemployment measurement tools.

Retraining programme: Government or corporate-funded initiatives designed to help displaced workers acquire new skills for different roles. Useful in specific, limited circumstances; structurally inadequate as the primary policy response to AI-scale displacement.

Productivity gain taxation: The principle that companies generating economic value through automation should contribute to the social costs of the displacement that automation produces, through tax mechanisms applied to capital returns rather than labour income.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

The next essay examines the political failure in full: why the left has a ready-made narrative that fits the crisis even if its proposals are inadequate, why the right has no coherent response at all, and what a serious cross-ideological politics of the AI transition would actually require.

If this essay reached you through someone else, consider subscribing directly to The Human Transition on Substack. The series continues weekly.

Sources and further reading:

• Brookings Institution — “New Data Show No AI Jobs Apocalypse — For Now,” February 2026

• Economic Innovation Group — “AI and Jobs: The Final Word,” August 2025

• Federal Reserve Bank of Boston — “Gig Workers Are Undercounted — or Unseen,” 2024

• Federal Reserve Bank of St. Louis — “Is AI Contributing to Rising Unemployment? Evidence from Occupational Variation,” August 2025

• ScienceDirect — “Artificial Intelligence and Technological Unemployment,” July 2025

• Upwork Future Workforce Index, 2025

• SQ Magazine — Freelance Economy Statistics 2026

• Self-Employed.com — “AI Is Flooding Freelance Platforms,” April 2026

• Fiverr / Purposed / Sensor Tower — platform download data, 2024–2026

Subscribe now

Elon Musk has a vision of the future that he describes, without irony, as something that “kind of sounds like heaven.”

In this future, AI and robotics have eliminated scarcity. Everyone has excellent medical care. Everyone has whatever goods and services they need. Work is optional — a hobby, like playing a video game, pursued for pleasure rather than survival. Saving for retirement is pointless because, as Musk told the Moonshots with Peter Diamandis podcast in January 2026, “It won’t matter.” He has said, on multiple occasions and in multiple forums, that AI and robotics are “the only way to make everyone wealthy” — and that he is “doing his best to make this happen.”

Sam Altman, CEO of OpenAI, has said he can “easily imagine a world where 30 to 40 percent of the tasks that happen in the economy today get done by AI in the not very distant future.” Dario Amodei of Anthropic has suggested that AI will beat humans “in almost everything” by 2027. At Davos in January 2026, the consensus among AI’s most powerful builders was not whether transformative disruption is coming — but how fast.

These are not fringe predictions. These are the stated beliefs of the people who control the most significant technology companies in the world, whose products are already deployed at scale in every major economy, and whose decisions will do more to shape the next fifty years than any elected government currently in office.

So let us take them seriously — not by accepting them uncritically, but by asking the question they consistently fail to answer.

If all of this is true, how do we get from here to there? And what happens to the billions of people in between?

Why the Bridge Is the Problem

The destination Musk describes — universal abundance, optional work, no poverty — is not inherently undesirable. The question is not whether the destination is appealing. The question is whether the bridge to it exists, who controls it, who pays for its construction, and what happens to the people who fall off it while it is being built.

This is not a hypothetical concern. The transition is already underway. In 2025, nearly 55,000 job cuts in the United States were directly attributed to AI, according to Challenger, Gray & Christmas, out of a total 1.17 million layoffs — the highest level since the 2020 pandemic. Major companies explicitly cited AI when announcing cuts: Workday eliminated 8.5 percent of its workforce to reallocate resources toward AI investment; Amazon cut 14,000 corporate roles, stating that AI enables leaner structures and faster innovation.

McKinsey research from late 2025 estimated that today’s technology — not future iterations, but what exists right now — could, in theory, automate approximately 57 percent of current U.S. work hours. That is not 57 percent of jobs disappearing tomorrow. It means that more than half of the working hours currently performed by human beings involve tasks that existing AI systems are technically capable of handling. Deployment is the limiting factor, not capability. And deployment is accelerating.

By the end of 2026, 37 percent of business leaders plan to replace workers with AI. Goldman Sachs estimates 300 million jobs globally will be impacted by 2028.

The people losing jobs today are not waiting for the abundance to arrive. They are living in the gap — the transition period between the world being dismantled and the world being promised. And that gap has no scheduled end date, no management plan, and no person or institution responsible for the people inside it.

This is the central problem that no abundance narrative addresses. And it has four possible resolutions — four futures that the current trajectory could produce. None of them are certain. All of them are live possibilities. And understanding which one you are currently heading toward is the most important piece of analysis available to any person trying to make decisions about their life, their family, or their community right now.

Future One: The Abundance Promise — If It Works

The most optimistic scenario takes the AI builders at their word and assumes their promises translate into reality on the timeline they suggest.

In this future, AI and robotics do generate enormous productivity growth. Goods and services become dramatically cheaper. Medical advances extend and improve life. Energy becomes nearly free as AI optimises both production and consumption. The economy is so productive that some form of wealth distribution — whether Universal Basic Income, a sovereign wealth fund, or a direct dividend from AI companies — provides every person with a meaningful standard of living regardless of whether they work.

This is not science fiction. Versions of this argument are made by serious economists, not just tech billionaires. Sam Altman has proposed what he calls the “American Equity Fund,” a mechanism where large AI companies and landholders would contribute a percentage of their value annually to a fund distributed to all citizens — effectively transferring a share of ownership of the automated economy to the populace.

The problems with this scenario are not technological. The technology may well deliver what is promised. The problems are political and structural.

Even $1,000 monthly payments to every American would cost $3.7 trillion annually — 50 percent of today’s federal budget. No political coalition currently exists to fund this. Our tax code currently rewards companies for replacing workers — subsidising AI investment while taxing the wages of the people being displaced. Reversing that structure would require a level of political will that is entirely absent from any major democracy at this moment.

More fundamentally: the abundance scenario requires the companies generating the wealth to agree to distribute it. There is no mechanism — legal, political, or market-based — that currently compels them to do so. Altman’s American Equity Fund is a proposal, not a policy. Musk’s “universal high income” is a stated aspiration, not a commitment. And both men are primarily accountable to their shareholders and investors, not to the workers whose income their products are displacing.

The abundance future is possible. It is not the default and, again, the bridge from here to there is far from being designed, let alone built.

Future Two: Corporate Feudalism — The Direction We Are Currently Moving

The second scenario does not require pessimism to describe. It requires only an honest reading of the current trajectory.

In this future, the companies that build and control AI accumulate a form of economic power that makes democratic governance structurally irrelevant. Not through a coup or a conspiracy — through the ordinary operation of the market. They own the most productive assets in the economy. They control the infrastructure that everyone depends on. They employ the workers whose skills remain valuable. Governments depend on their tax revenues, their investment decisions, and their technological capabilities.

In the last quarter of 2025, the largest technology companies spent $142 billion on AI infrastructure — chips, data centres, power systems — in a single quarter. That is a level of capital deployment that no legislature, in any democracy, is currently matching with investment in the public infrastructure, retraining programmes, or social safety nets that a displaced workforce would require.

The outcome of this trajectory is not a dramatic takeover. It is a gradual reorganisation of power. States retain their formal authority — their constitutions, their parliaments, their elections — while their practical ability to make decisions that constrain major technology companies erodes. The companies do not need to control governments. They merely need to be indispensable to them.

This is not unprecedented. It is, in fact, the structure that has historically emerged when one entity controls the dominant means of production in a region or economy. Company towns in the industrial era are one version of this. The dependency of entire national economies on single commodity exporters is another. The difference now is the scale — global rather than regional — and the speed at which it is consolidating.

63 percent of American workers currently believe AI will decrease overall job availability. The institutional projections are more conservative, but the lived experience of workers — falling wages, contracting opportunity, the gradual sense that the rules of the economy are being rewritten without their input — is already registering.

This future does not announce itself. It arrives through a series of decisions that each seem individually reasonable: a company reduces headcount because AI makes the work more efficient; a government defers regulation because the technology is moving too fast to govern effectively; a pension fund invests in AI infrastructure because the returns are compelling. Individually, none of these decisions looks like a civilisational choice. Collectively, they are.

Future Three: State Control — The Overcorrection

The third scenario is the reaction to the second. Governments, faced with mass unemployment, collapsing tax bases, and social instability, intervene forcefully. They nationalise AI infrastructure. They impose robot taxes. They regulate deployment so heavily that the pace of automation slows dramatically. They redistribute the gains from AI productivity through expanded welfare systems, state employment programmes, and direct wealth transfers.

This scenario is the one most likely to appeal to people who are alarmed by Future Two. It sounds like a correction. In many respects, it is. But it carries its own dangers, which the twentieth century illustrated with considerable precision.

The danger is not that governments will intervene in the economy. Democratic governments have always done this, and the argument that they should continues to carry weight. The danger is the concentration of economic control that intervention at this scale would require.

If a state controls the AI systems that manage healthcare, education, transport, energy, food production, and financial services — which is what controlling “the means of production” would mean in an AI economy — then the state has a form of power over individual life that is qualitatively different from any welfare state that has previously existed. The USSR comparison is not hysteria. It is a reference to what actually happens when one entity controls the economic infrastructure on which everyone depends: inefficiency, yes, but also surveillance, conformity, and the systematic subordination of individual freedom to institutional priority.

This does not mean state intervention is necessarily wrong in every situation. It means that the form, scale, and governance structure of that intervention matters enormously — and that the political systems of most democracies are not currently designed to manage an intervention of this nature without the corruption of the democratic process itself.

The third future, pursued without extraordinary institutional care, risks exchanging corporate feudalism for state feudalism. The master changes. The structure of dependence does not.

Future Four: Collapse — The Failure to Manage the Bridge

The fourth scenario is not a dramatic end. It is a prolonged unravelling.

In this future, none of the above gets managed in time. The political will for redistribution does not materialise. The corporate accumulation of power continues unchecked. The social systems built for a labour economy — pensions, welfare, healthcare funding, education — face fiscal crises they cannot resolve because the tax base they were designed for is eroding. Inequality reaches levels that historical experience suggests are incompatible with stable democratic governance.

The result is not a single catastrophic event. It is a cascade: rising unemployment, then social unrest, then political instability, then the erosion of democratic norms, then the capture of weakened institutions by authoritarian movements that offer simple answers to genuinely complex problems. This pattern is not new. It has appeared, in various forms, every time an economic transformation has outrun the capacity of political institutions to manage its consequences.

The National Bureau of Economic Research’s 2025 occupational analysis found that approximately 3.9 percent of U.S. workers — roughly 5 to 6 million people — sit at the intersection of high AI exposure and low adaptive capacity: those in routine roles, with limited savings, in labour markets with fewer alternative job options. This is where the genuine hardship concentrates. Five to six million people in one country alone, in the early stages of a transition whose scale is still accelerating.

The collapse scenario does not require a majority of workers to be displaced. It requires enough displacement, concentrated in enough communities, to generate the political volatility that historical experience shows is the precursor to democratic backsliding. The Weimar Republic was not destroyed by mass starvation. It was destroyed by enough economic insecurity, concentrated in enough places, to make a politics of resentment and scapegoating more compelling than the ordinary machinery of democratic governance.

Where We Are Currently Heading

The honest answer is: toward a combination of Futures Two and Four, with the hyped possibility of Future One as a destination that nobody has designed or built the road to.

The technology is advancing faster than the political institutions designed to govern it. The wealth it is generating is concentrating in fewer hands faster than any redistribution mechanism is being designed to address. The workers being displaced today are being displaced now — not in a managed transition with retraining programmes and safety nets, but through ordinary layoffs, falling wages, and contracting opportunity. The political debate about what to do is happening, in most democracies, at a pace and depth entirely disproportionate to the scale of the challenge.

Musk told the world in January 2026 not to worry about saving for retirement in 10 or 20 years — because AI will have made money irrelevant. That advice is directed at people who still have 10 or 20 years before they retire. It is cold comfort to the 52-year-old logistics manager whose role was eliminated last year, the 34-year-old graphic designer whose client base has evaporated, or the 45-year-old paralegal whose firm has reduced its junior staff by half since 2023.

For those people, the future is now. The bridge is what they are standing on. And it is being built — or not built — by people who will not be standing on it themselves.

What This Means for How You Make Decisions

Understanding these four futures is not an abstract intellectual exercise. It is a practical necessity for anyone making decisions about their career, their finances, their children’s education, or their community’s future.

Several things follow clearly from this map.

The abundance future requires advocacy, not patience. It will not arrive by default. It requires political pressure for redistribution mechanisms, for AI taxation, for democratic governance of transformative technology. People who want this future need to demand it — from their political representatives, from the companies whose products they use, and from the public conversation about what the economy is for.

The corporate feudalism future is the path of least resistance. It requires nothing from anyone. It simply requires the current trajectory to continue. Recognising it means watching for the specific signs: declining labour share of income, consolidation of AI infrastructure in fewer hands, the quiet erosion of regulatory capacity, the growing gap between productivity and wages.

The state control future needs to be designed carefully or not at all. The instinct to have governments intervene is correct. The design of that intervention — its governance structures, its accountability mechanisms, its limits — determines whether it produces a social democracy or a surveillance state. These are not the same thing, and the difference between them is not guaranteed by good intentions.

The collapse future is preventable but not inevitable. It requires enough people to understand what is happening — and to demand, before the institutional capacity to respond has been eroded, the kind of serious public conversation and political action that the scale of the transition requires.

This platform exists to contribute to that understanding. Not to predict the future — nobody can do that honestly. But to map the territory as clearly as the evidence allows, so that the people navigating it can do so with open eyes rather than the false comfort of promises nobody has the structural ability to keep.

The train is moving. The destination is not fixed. And the decisions made in the next five years — by governments, by companies, by organised workers, and by ordinary citizens who understand what is at stake — will do more to determine which future arrives than any technology built in any laboratory.

Key Terms Explained

Abundance scenario: The prediction, advanced most prominently by Elon Musk and other AI optimists, that AI and robotics will generate sufficient economic productivity to provide universal material security, making traditional employment unnecessary.

Corporate feudalism: A term used to describe a social and economic structure in which large private corporations, rather than governments or markets, become the primary organisers of economic life — a potential outcome of unchecked concentration of AI-generated wealth and power.

Universal Basic Income (UBI): A policy proposal in which all citizens receive regular, unconditional cash payments from the government, regardless of employment status, designed to provide a basic standard of living as automation reduces the demand for human labour.

AI governance: The set of policies, regulations, international agreements, and institutional mechanisms designed to ensure that AI development serves the public interest and is subject to democratic accountability.

Labor displacement: The process by which workers lose employment due to technological automation — as distinct from cyclical unemployment (caused by economic downturns) or frictional unemployment (caused by the time taken to match workers to new roles).

The transition gap: The period between the current economy, which depends on human labour for its tax base, welfare systems, and social organisation, and any future economy in which AI-generated productivity is distributed broadly enough to sustain those systems without mass human employment.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

The next essay in this series examines the specific professions already being displaced — not in ten years, but now. If you want to receive it directly, subscribe to The Human Transition on Substack.

Sources and further reading:

• Goldman Sachs Global Investment Research — AI and the Labour Market (2023, updated 2025)

• McKinsey Global Institute — State of AI 2025

• World Economic Forum — Future of Jobs Report 2025

• National Bureau of Economic Research — Occupational Analysis and AI Exposure (2025)

• Challenger, Gray & Christmas — Job Cuts Report 2025

• Elon Musk, Moonshots with Peter Diamandis podcast, January 2026

• Sam Altman, Die Welt interview, 2025

Subscribe now

Yesterday, something happened at the Vatican that most technology journalists covered as a curiosity and most religion journalists covered as a theological event. Neither framing is adequate.

Pope Leo XIV released his first encyclical — a formal teaching document, the most authoritative form of papal writing — and it was entirely about artificial intelligence. It is titled Magnifica Humanitas, which translates as “Magnificent Humanity.” The document runs to 42,300 words in its English version, roughly the length of a short book. It addresses AI governance, the dignity of work, the displacement of workers, lethal autonomous weapons, transhumanism, and what it calls the temptation to build a future that excludes both God and human beings.

And standing on the stage at the Vatican press conference, alongside the Pope, was Christopher Olah — co-founder of Anthropic, the AI company behind the Claude assistant, currently valued at approximately $900 billion.

This is not a religious story. It is not a technology story. It is a story about power, about accountability, and about the almost complete absence of any serious public reckoning with what is being built and who it is being built for. And it is a story that touches every person reading this — regardless of whether they have ever set foot inside a church.

Why an AI Company Was on Stage at the Vatican

Let us start with the obvious question: what was Anthropic doing there?

The relationship between Anthropic and the Vatican did not materialise overnight. It sits within a multi-year conversation that the Catholic Church has been having, quietly and seriously, about the human consequences of artificial intelligence. Under Pope Francis, the Vatican initiated the Minerva Dialogues in 2016 — a forum bringing together theologians, ethicists, and technology professionals to wrestle with exactly the questions that mainstream political discourse has largely avoided.

Anthropic was invited because, among the major AI companies, it has positioned itself most explicitly as a safety-focused organisation. Its co-founders, Dario and Daniela Amodei, left OpenAI specifically over concerns about the pace and recklessness of AI development. The company’s very name is derived from the word “anthropic” — relating to human existence. It has publicly refused to develop lethal autonomous weapons. It delayed the release of a model in February 2025 after testing indicated it could be used to help develop biological weapons.

Some observers have described Anthropic’s Vatican presence as a calculated business move — gaining moral credibility at a moment when the company is at odds with the Trump administration, and seeking market footholds in European nations where regulatory sentiment is more cautious. That reading is probably not wrong.

But dismissing the presence of an AI company at a papal encyclical launch as purely transactional misses something important. It tells us that the Vatican — one of the most ancient and strategically sophisticated institutions in human history — has concluded that the AI transition is serious enough to require direct, public, formal engagement with the companies driving it. That is itself a signal worth paying attention to.

What Magnifica Humanitas Actually Says

The encyclical was signed on May 15, 2026 — a date chosen deliberately. It is the 135th anniversary of Rerum Novarum, Pope Leo XIII’s landmark 1891 encyclical on workers’ rights, written in direct response to the upheaval of the Industrial Revolution. The current Pope chose the name Leo XIV for the same reason: to position himself explicitly in that tradition, as the pope who addresses the rights and dignity of workers in an age of technological transformation.

The parallel is precise and intentional. In 1891, industrial capitalism was reorganising the economy at a speed that democratic and social institutions could not match. Workers were being displaced, exploited, and stripped of the security they had built their lives around — without any plan, without any compensation, and without any voice in the decisions being made. Rerum Novarum was the Catholic Church’s formal intervention in that crisis — a document that influenced labour law, trade union rights, and social policy across the Western world for the following century.

Pope Leo XIV is making the same move, 135 years later. The crisis is different in its technology. It is identical in its structure.

Magnifica Humanitas opens with a framing that is stark even by the standards of encyclicals: “Humanity, created by God in all its grandeur, is today facing a pivotal choice: either to construct a new Tower of Babel or to build the city in which God and humanity dwell together.”

The Tower of Babel reference is not mere religious decoration. In the biblical account, the Tower of Babel represents humanity’s attempt to build something that reaches beyond its own limits — a project of pure technological ambition, divorced from ethical constraint, that ultimately fragments and destroys the community that built it. The Pope is saying, in the most direct language available to him: this is the risk we are currently running.

The document then develops its argument across five chapters, moving from a diagnosis of the current moment, through the Church’s social doctrine on human dignity and work, to specific prescriptions for AI governance, workers’ rights, and the ethics of autonomous weapons.

For our purposes — for anyone who is not a theologian and is trying to understand what this document means for their life — several passages stand out.

On the Dignity of Work — and What Automation Threatens

The encyclical is unambiguous on the question that most political discourse carefully avoids. Work is not simply a means of generating income. It is, in the Catholic social tradition, a fundamental good — a way in which human beings express their freedom, creativity, and capacity for cooperation. To deprive a person of meaningful work is not merely an economic problem. It is a violation of their dignity.

Directly addressing automation, the document states that job insecurity, fragmented career paths, and the displacement caused by AI “must not be evaluated solely in terms of efficiency, but in relation to the dignity of the worker, the right to sufficient remuneration and the genuine possibility of participating in society.”

This is a direct challenge to the dominant framing in economic and political debate, which evaluates automation almost entirely through the lens of productivity and growth. The encyclical insists on a different measure: what does it do to the person?

The document also addresses what it calls “structures of sin” — drawing on the language of Pope John Paul II — to describe economic and political arrangements that systematically marginalise people and exclude them from the benefits of development. The encyclical is explicit: no person or people should be treated as expendable in the processes of development. That phrase — expendable — is not a theological abstraction. It is a precise description of what happens when a company automates a role, releases the workers, reports efficiency gains to shareholders, and moves on.

On Technology, Power, and Accountability

The encyclical contains a line that should be read carefully by anyone thinking seriously about the AI transition: “Technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate, and use it.”

This is the sentence that renders naive any discussion of AI as simply a tool. A tool is neutral. A hammer does not have characteristics. But AI systems are shaped by the priorities, incentives, blind spots, and commercial interests of the organisations that build them. They encode assumptions. They embed values. They reflect — and reinforce — the power structures of the world in which they are created.

The encyclical does not name specific companies. It does not need to. The argument is structural: when a small number of private entities control a transformative technology, and when those entities are primarily accountable to their shareholders rather than to any democratic institution, the question of who benefits — and who pays the cost — is not a neutral technical question. It is a political and moral one.

Pope Leo is not the first person to make this argument. But he may be the first person with the institutional reach and moral authority to make it in a way that cannot be easily dismissed as technical ignorance, political partisanship, or anti-progress hysteria. This is, after all, the head of a church with 1.4 billion members, publishing a formal teaching document that has been in preparation for years, drawing on the work of ethicists, theologians, technologists, and social scientists.

Why Anthropic’s Presence Raises More Questions Than It Answers

Here is where the story becomes more complex — and more revealing.

Anthropic’s presence on the Vatican stage is, in one sense, reassuring. It suggests that at least one significant AI company is willing to engage seriously with the moral and ethical dimensions of what it is building. Dario Amodei has spoken publicly about his belief that AI development carries genuine existential risks. Anthropic’s research on AI safety and alignment is taken seriously within the field. The company has, at various points, prioritised caution over speed in ways that its competitors have not.

But the presence of a $900 billion company at the launch of a papal encyclical on AI ethics also raises a question that the encyclical itself gestures toward: who is being held accountable, and by whom?

An encyclical is a moral argument. It has no enforcement mechanism. The Vatican cannot compel Anthropic, or any other AI company, to do anything. The companies that choose to engage with the Catholic Church’s social teaching do so voluntarily — which means they also choose the terms of that engagement. And a company that participates in a dialogue about AI ethics while simultaneously deploying AI systems that displace workers, at scale, across every sector of the global economy, is not thereby absolved of the consequences.

The encyclical itself acknowledges this tension. It calls for genuine democratic oversight of AI, for international governance mechanisms, for the protection of workers displaced by automation, and for rigorous ethical constraints on autonomous weapons systems. These are not things that any individual company — however well-intentioned — can deliver on its own. They require the political will of states, the coordination of international institutions, and the engagement of democratic publics that currently have almost no meaningful voice in the decisions being made.

That political will does not currently exist. That international coordination does not currently exist. That democratic engagement is almost entirely absent.

The Question Nobody Is Asking

The mainstream coverage of Magnifica Humanitas has focused on the religious dimensions: the Pope’s theological framework, the Catholic social teaching tradition, the question of what it means for AI to respect human dignity in a spiritual sense.

That is a legitimate conversation. But it is not the conversation that most people need to have right now.

The more urgent conversation is the one the encyclical points toward but that secular politics has almost entirely abandoned: how do we govern a transformation of this magnitude, at this speed, in a world where the institutions designed for democratic governance are structurally unable to keep up?

Think about who is currently making the decisions that will define the economic and social landscape for the next fifty years. It is not elected governments. It is not international bodies with democratic mandates. It is a small number of private companies — American, primarily — whose primary accountability is to their investors. They are moving at a speed that regulatory frameworks cannot match, in a direction that democratic populations have never been asked to approve, with consequences that will affect every person on earth.

The teacher whose students are wondering whether their degree is worth pursuing. The accountant whose firm has quietly reduced headcount by 30 percent in two years. The logistics worker whose warehouse is being automated. The freelancer whose rates have been falling since 2023. The parent who does not know what to advise their child about education and career. The future pensioner who does not know whether the systems they have paid into for decades will still exist when they retire.

None of these people voted for this. None of them were consulted. None of them have a seat at the table where these decisions are being made.

Pope Leo XIV knows this. It is precisely why he chose to sign his encyclical on the anniversary of Rerum Novarum. The workers of 1891 did not vote for the Industrial Revolution either. The question then, as now, was whether the institutions of society — religious, political, and civic — would intervene on their behalf, or whether they would be left to absorb the costs of a transformation driven entirely by the interests of those with the capital and the technology.

What This Means for You — Regardless of Your Faith

You do not need to be Catholic to understand why this encyclical matters. You do not need to be religious at all.

What the publication of Magnifica Humanitas tells us — with the presence of an Anthropic co-founder on the stage — is that the conversation about AI is no longer confined to technology conferences, venture capital boardrooms, and specialist academic journals. It has reached the Vatican. It has reached the institution that, for better or worse, represents the moral conscience of 1.4 billion people.

That matters because it changes the landscape of accountability. When the Pope publishes a 42,000-word document arguing that the displacement of workers by AI is a moral issue, not merely an economic one, it becomes harder to dismiss that concern as technophobia or anti-progress sentiment. When he invokes the language of dignity, of structures of sin, of the common good, he is deploying a moral vocabulary with centuries of social and political weight behind it.

It also matters because of what it signals about the seriousness of the moment. The Vatican does not publish encyclicals lightly. The preparation of Magnifica Humanitas took years, involved dozens of consultations, and drew on the work of theologians, ethicists, economists, and technologists. The decision to publish it as Pope Leo XIV’s first encyclical — before anything else — tells us that the Church regards this as the defining moral challenge of the current era.

And the question it raises — the question that this publication, this platform, and every piece of work we do here is organised around — remains unanswered by any of the people with the power to answer it.

How do we get from here to there?

How do we navigate the transition between a world built on the assumption of human economic value and whatever comes next, without the kind of catastrophic social breakdown that has historically accompanied transformations of this magnitude? Who is responsible for the workers displaced by automation? Who will fund the social systems that were designed for a labour economy? Who will hold the companies accountable? Who will ask them the hard questions — and who will have the power to enforce the answers?

The Pope asked. The AI company stood on the stage and listened. But the encyclical has no enforcement mechanism. Anthropic’s presence changes the optics of the situation without changing its structure.

The train is still moving. The questions are still unanswered. And most of us are still picking up pennies on the track.

Key Terms Explained

Encyclical: A formal letter issued by the Pope, addressed to the entire Catholic Church and to “all people of goodwill.” Encyclicals represent the highest level of ordinary papal teaching and carry significant weight in Catholic doctrine.

Rerum Novarum: Pope Leo XIII’s 1891 encyclical on the rights of workers and the conditions of labour during the Industrial Revolution. It is considered the founding document of Catholic social teaching on economics and is credited with influencing labour law across Europe and the Americas throughout the twentieth century.

Magnifica Humanitas: Latin for “Magnificent Humanity.” The title of Pope Leo XIV’s AI encyclical, signed May 15, 2026 and released May 25, 2026. The document addresses AI governance, workers’ rights, human dignity, transhumanism, and autonomous weapons.

Anthropic: A San Francisco-based AI research and development company, co-founded in 2021 by Dario and Daniela Amodei and other former OpenAI employees. Creator of the Claude AI assistant. Currently valued at approximately $900 billion.

AI alignment: The field of research concerned with ensuring that AI systems behave in ways consistent with human values and intentions. One of the central concerns of Anthropic’s research programme.

The Human Transition: Navigating the Age of AI Casualties is an independent platform covering the human consequences of AI — for the people living through the transition, not the people building it. Written by Miguel Aparicio.

If this essay was useful to you, consider sharing it with someone who needs to understand what is happening — regardless of whether they follow technology news.

Subscribe now