Payback’s a Bitch: AI Sovereign Wealth Funds, the Fake Andrew Keen, and America’s Inevitable Decline

“The frontier AI companies invited the government into the room. Now the government is beginning to behave as if it owns the door, the guest list, the schedule, and the product roadmap.” — Keith Teare

Last week, I was away in Europe. So Keith Teare ran our That Was The Week show solo — with a chillingly authentic Andrew Keen bot. So realistic, in fact, that the fake version sounds (to me, at least) more interesting than the real one.

The bad news is that I’m back. The good news is it’s been an interesting week in tech. That was the week in which the US Commerce Department told both OpenAI and Anthropic that they now need government approval for whom they can sell their frontier AI models. This is supposedly “voluntary” — for now, at least.

Keith’s TWTW editorial argues that Dario Amodei and Sam Altman have spent over a year crying wolf about the dangers of their own technology, supposedly deliberately seeking government involvement as a regulatory moat against competitors. And now the government has walked through the door that Sam and Dario left ajar. Now, Keith argues, the US government is behaving as if it owns not just the door and the guest list, but the entire product roadmap. “Payback’s a bitch,” Keith bristles in his editorial.

The other major news this week is the rumour (via David Sacks) that OpenAI has offered the US government a 50% stake in a sovereign wealth fund. If true, this would change everything — not just in Silicon Valley, but in the political debate about public ownership of our AI economy. It’s not just tech insiders like Sacks and Altman who are on board the sovereign wealth fund express, but also Bernie Sanders and other leftist critics of Big Tech. So maybe payback, at least when it comes to public investment in AI, isn’t always such a bitch.

Five Takeaways

•       The Fake Andrew Keen: An Hour of Work on a Local Nvidia Card: Keith ran last week’s show solo with an AI-generated Andrew Keen: trained on a few episodes of the show, animated from a YouTube still, scripted from Keith’s newsletter. No third-party service. Just a local PC with an Nvidia GPU, about an hour of work, three attempts. Andrew, listening back, second-guessed whether he was actually there. The result was “pretty bad compared to our normal actual live shows,” Keith says. But also: really good. The question hanging over this episode and every future one: which Andrew are you listening to?

•       Payback’s a Bitch: How AI Companies Created Their Own Regulatory Trap: The US Commerce Department has told OpenAI and Anthropic they need government permission for who gets to use their latest models. Voluntary, for now. Keith’s diagnosis: AI leadership spent more than a year crying wolf about existential risk — not because they believed it, but because government regulation creates a moat against competitors. Now the government has taken them at their word. Dario and Sam Altman wanted to be wrapped in government clothing. They are. The government now owns the door. They asked for this. They got it.

•       American and Chinese State Capitalism: Converging Models: Andrew raises the macro argument: what we’re watching is the convergence of American and Chinese models of capitalism toward a more state-centric model. China has always been explicit about state control. America has prided itself on free enterprise — even when the internet, atomic technology, and now AI were all substantially government-funded or government-shaped. Keith agrees at this level: all governments seek to control things they frame as dangerous. The difference is the framing. The direction of travel is the same.

•       OpenAI’s Rumoured 50% Stake Offer to the Government: Keith has heard — fro