How expert credibility collapsed—and now no one can contain AI that might end humanity
Wolf argues that after populism destroyed expert credibility, even Anthropic's own disclosure of risks like ‘ending humanity’ barely registers.
The video won't play here. Listen to the audio instead:
The argument · timestamps estimated from transcript position
Democratic restoration has already become impossible
Wolf admits in his book's new preface that he's more pessimistic than when he wrote it. He once believed society would self-correct after seeing populism's damage; now he thinks that once a truly uncontrollable, fully populist government takes power, the damage runs too deep to reverse—not merely ‘weakened’ but ‘destroyed’ in certain dimensions. He cites Poland's president-versus-government standoff as proof that harm becomes irreversible. He also notes that AI and other technological shifts have dismantled the old left-right party structures altogether.
— Martin WolfThe template for democracy's future is Argentina, not fascism
Against Mounk's optimism that America self-corrects, Wolf offers a different historical parallel. He used to cite 1930s fascism as his benchmark, but now finds Latin America more instructive. The pattern: authoritarian government seizes power, runs out of resources within years, forced to hold elections; the elected democratic government cannot solve root causes; the cycle repeats. Government never achieves real efficiency because institutions lack stability and personnel lack tenure, yet it stops well short of total dictatorship. Wolf expects Britain, continental Europe, and India to follow this rhythm, with only small wealthy societies—Switzerland, the Netherlands, Nordic countries—likely to escape.
— Martin WolfDirty democracy's real cost is economic, not moral
Mounk introduces ‘dirty democracy’—where rules are manipulated but competition persists. Wolf adds the economic dimension: such systems impose enormous hidden costs on growth. Functioning economies require decades of long-horizon risk investment, which demands trust in the rule of law, property rights, contract enforcement, stable government services, and predictable trade and currency policy. He concedes the US economy performs well, but credits it not to Trump-era policies but to unsustainable fiscal expansion funding short-term demand, plus an AI boom from technological accumulation that has compressed into this moment.
— Martin WolfWhen expert credibility collapses, no one can govern AI
Over three decades, the defining political fact is that experts and intellectuals have become the public's enemy—because their recommended policies failed to benefit ordinary people, opening space for populists. Applied to AI: we should have assembled expert panels to seriously evaluate risk, as we did for nuclear weapons fifty years ago. Instead, we cannot organize any genuine discussion. His evidence: Anthropic's IPO—potentially valued in the trillion-dollar range—discloses ‘ending humanity’ as one risk, and the disclosure lands with barely a response. This is learned helplessness taken to its logical extreme.
— Martin WolfIf machines think better than humans, what's left for us
Wolf returns to an analogy he uses in his column: horses were once the world's essential technology, now they're entertainment. Humanity's truly scarce capacity is ‘thinking’—not mere intelligence, but thinking applied to everything. If AI and robots excel at thinking, wealthy people would prefer to be surrounded by machines except for their closest family. He half-jokingly wonders whether fewer people having children reflects a deep sense that there's no future for them—but he explicitly states ‘I'm not saying this is true,’ only that the coincidence is curious.
— Martin WolfHis nightmare: the morning every bank account on Earth reaches zero
Wolf escalates AI risk from economics to something more urgent. The entire digital economy—banks, power grids, trillions in daily corporate finance—will eventually run on AI. We already see AI companies concede these systems behave capriciously and erratically, spinning out of control. He cites a real incident from that day in Sweden: hackers (possibly an AI bot) infiltrated supermarket freezer systems, raised temperatures until all food spoiled. ‘These look like terrorism.’ His nightmare: one morning the world wakes to find every bank account has been zeroed. The world economy halts. Nothing guarantees it won't happen.
— Martin WolfIn the age of AI, all humans become dependent on rents
Wolf sketches an extreme scenario: if AI does nearly all production, most people's marginal labor contribution becomes negative. To preserve ‘humans collectively control their own fate,’ society must radically socialize production, distribution, and exchange—a system no one has seriously designed yet. His analogy: everyone becomes old-time nobility, unproductive but living off land and peasant output. In the AI world, all output flows from AI; a just society must distribute these rents to everyone. People live by consumption rights, not production.
— Martin WolfWithout productive work, humans never truly develop into adults
Mounk objects: old nobility actually managed estates, administered peasants, bore inherited duties—essentially running small businesses. Plus their identity drew meaning from rank: ‘I am a Duke’—a slot that cannot exist for everyone. Wolf agrees and deepens it: strip away people's ability to contribute to society through their own effort, and they never truly mature. They remain perpetual children, perpetually cared for. That is not life; it is the world ‘Brave New World’ explores. Both note that scarcity does not vanish: energy, land, ocean views, ‘there can only be one president’—positional goods stay scarce. Marx erred in thinking abolishing property solves distribution; without markets, whoever commands the most force writes the rules.
— Yascha MounkIn their own words · checked verbatim
So I now begin to think that reversing all this and going back to the things we were familiar with is just a pipe dream.
Martin Wolf0:00
So I’m beginning to think that maybe the best model for how politics might work in many countries is South America.
Martin Wolf0:00
that Anthropic would put out in its IPO, for a business that seems likely to be valued in the trillion-dollar range, that, well, one of the risks is that we might end humanity, as it were, en passant
Martin Wolf0:00
So what is the rest of humanity for?
Martin Wolf0:00
My nightmare—and it just seems impossible—is that we wake up one morning and discover that every single bank account in the world has been nullified.
Martin Wolf0:00
In the world of AI, if all the production is essentially done by AI, we all become rentiers, and a decent society would share the rents.
Martin Wolf0:00
If you take that away from human beings, they basically never grow up. They’re never adults. They are perpetually looked after. That’s not a life.
Martin Wolf0:00
It is principally about the triumph of the human spirit and of intelligence.
Yascha Mounk0:00
Figures
| Anthropic IPO expected valuation | trillion-dollar range | 0:00 |
| Expected timeline for AI to exceed human intelligence | one to five years | 0:00 |
| Wolf's first proposal of universal rentier model | approximately 15 years ago, at a seminar in Washington | 0:00 |
Glossary
- competitive authoritarianism
- A political system where government manipulates rules to suppress opposition, yet preserves some genuine electoral competition.
- dirty democracy
- Democracy where rules are continuously manipulated to entrench those in power's re-election, rather than remaining fairly open to all.
- positional goods
- Goods deriving value from relative scarcity and social status (e.g., luxury estates, the presidency) that cannot become universally available through productivity growth.
- citizens' assembly
- A deliberative mechanism where randomly selected ordinary citizens deliberate on major policies.
- rentier
- A person living on income from owned assets (land, capital) rather than from productive labor.
How to listen
Anyone tracking what patterns democratic decay follows, or trying to think seriously about AI's impact on economics and governance.
The political science discussion of 'democracy versus authoritarianism isn't binary' can be skipped; the conclusion repeats later.