I SUPPOSE IT’Smy own fault for not making this part of the headline or subheadline in my Zeteo column on Friday:Never would I ever call for bailing out the artificial-intelligence oligarchs. But once Zeteo‘s Instagram account posted my column—a column I wrote in the hope of expanding the ambitions of the American left for expropriating a deeply dangerous enterprise that is embedding itself within the Security State—I was hammered with comments misunderstanding my argument, or at least the version of it on the slideshow, as a call for a bailout. Nah.
I say it’s my own fault because I should have anticipated the criticism. This one strays a bit out of my usual lane. I ran the capsule version by a friend, someone who follows the artificial-intelligence industry with depth and skill far exceeding my own, whoalso initially thought I was advocating a bailout. I didn’t understand their reaction and unfortunately, deadlines being what they were, didn’t get clarity on how I could have differentiated my piece further. I’ll take my medicine here: It’s a pet peeve of mine when writers and journalists whine about being misunderstood. So this is on me. Thank you to Instagram user @raph_guerault for letting me know about the huge payments to industry that France made for its 1980s nationalizations. It’s my fault for not knowing about that, and not knowing that many readers would have such a model in mind for AI nationalization.
But no. The oligarchs wielding a weapon of mass destruction should get nothing, and consider themselves lucky to get it. This, as I argue in the piece, is a sovereignty question, and the people must win. The Defense Production Act can be the lever of victory.
This piece emerged from my brain being set on fire by a number of books I read this summer and strongly recommend. Those books are End Times Fascism by Naomi Klein and Astra Taylor, The Nerd Reich by Gil Duran, Muskism: A Guide For The Perplexed by Quinn Slobodian and Ben Tarnoff and Crack-Up Capitalism by Quinn. (Also, kind of, Jordan Harper’s novel A Violent Masterpiece.) None of these books proposed nationalization, and while I suppose I should have considered that a clue, I want a resurgent American left to get comfortable talking about taking industries away from their oligarchs. This piece is the result. I am going to put much more of it than usual outside the paywall in the hope of not being further misunderstood. But if I am, once again: that’s on me.
IN THE OPINIONof the artificial intelligence giant Anthropic, an AI model should be permitted tokill or seriously injure up to 50 people at a stretch.
A larger death toll than that, according to Anthropic’s scrupulous moral accounting, should prompt government regulators to block the release of a model. Similarly, Anthropic contends that frontier AI should get to risk causing as much as $1 billion of damage before regulators intervene. All that is contained in the draft of a bill Anthropic is pushing to shape the long-awaited contours of an AI regulatory framework. “These are sensible policies we believe Congress should act on,” a company spokespersontoldPolitico. Remember this the next time Anthropic positions itself as the humane AI company.
Not that it’s much comfort, but AI researchers, with astonishing casualness, project that their products will cull far more than 50 people at a time. Last month, Anthropic senior researcher Evan Hubinger tweeted that he considers AI to have a greater than 10% chance of “kill[ing] all humans… within the next decade.” Hubinger was reacting to Jacob Coxon, who announced he had quit Anthropic because it, and rival OpenAI, were “gambling with our lives.” At least Coxon, like Miles Dyson in ‘Terminator 2’, had the moral rigor to stop committing harm; and Sarah Connor didn’t even have to shoot him first. It is a portrait of capitalist nihilism that people like Hubinger would continue to develop a technology they believe might bring about extinction.
Coxon hastened to insist that AI researchers who make such predictions are not engaged in a “marketing stunt.” He said so because the handful of tech oligarchs who control AI are thoroughly untrustworthy. They haveinflated an AI bubble that threatens the U.S. economy. Many suspect that such apocalyptic forecasts are cynical, backdoor assurances of AI potency that puff the bubble up ahead of long-expected IPOs. The overwhelming majority of us, who have no visibility into the labs, have no choice but to wonder whether those inside of them are building an extinction machine or simply engaged in now-routine fraud. Maybe both?
All these should be considered clues.
The oligarchs who control the burgeoning AI industry are con men cosplaying as Oppenheimers. OpenAI’s Sam Altman, for instance, who like most of his colleagues is not a data scientist, was famously briefly fired by his board for being “not consistently candid in his communications.” In the hands of Altman and his few peers lies a technology whose widely anticipated use case, central to the economic expectations of AI’s champions, will be todecimate white-collar jobs, much as deindustrialization decimated blue-collar ones. Those peers exhibit a palpable sociopathy: Peter Thiel’sfamous long pause when asked if humanity should survive; or his Palantir colleague Alexander Karp’senthusiasm for Israel using his technology for its genocide; or the apparent Nazisalute of xAI’s Elon Musk, whose destruction of USAID is estimated to have led tothe deaths of over 750,000 people. Whether or not AI can murder us all, whether or not such predictions are valuation slop, the reality at this very moment is that AI is being used by the Israeli and U.S. militaries to generate targets at scale. The results areGaza and theMinab elementary school missile strike that killed more than 120 children. They will not be the last.
Whatever else AI is and whatever it might become, it is a weapon of literal and economic mass destruction. Anthropic’s IPO, released earlier this week, isexplicit about this: “[O]ur most capable models have demonstrated, and future models are expected to demonstrate, capabilities that could provide meaningful assistance to malicious actors seeking to develop chemical, biological, or other dangerous weapons.” It’s insane and unacceptable that weapons of mass destruction could be in private hands – let alone the hands of Epstein-class eminences likeJeffrey’s friend Peter. That’s nauseatingly ironic, considering the WMD pretexts for launching both the Iraq and Iran wars. The United States should nationalize the artificial intelligence industry before it kills at a greater scale. It might be a morbid symptom of our democracy that nationalization has yet to be treated as a serious option in the U.S. when such an obviously predatory elite can gamble, as Coxon put it, with the lives, wealth, and freedom of hundreds of millions of people.
THE BRAZENNESS OF ANTHROPICS’S regulatory model might distract from the absurdity, however normalized, of an industry proposing to regulate itself – which ought to be considered another clue. For years, a faction within the AI oligarchs, represented by Altman and Anthropic’s Dario Amodei, have called for government regulation. Often that has been portrayed as an indicator of their moral rigor and civic commitment. It is instead a sign that the industry knows that regulation poses no significant threat.
It’s a smart bet. Silicon Valley has thoroughly captured its regulators – a capture highlighted by tech and venture capital’sfuriousattacks on the former Federal Trade Commissioner Lina Khan, anotable exception to the capture rule. In the current U.S. congressional elections, super PACs aligned with Anthropic and OpenAI have poured a stunning$55.7 million into purchasing legislators who will exercise lax oversight and discipline regulators who might not.Most of their spending has gone to Democrats. Their xAI competitor, Elon Musk, has spent about five times more to elect Republicans – an absurd$276 million, which by his recent standards is a reduced amount of money to throw into politics.Thiel bankrolled the current Vice President JD Vance, and the sector will spend extravagantly on a Vance presidential run in 2028. A permissive regulatory environment doesn’t happen by accident.
But if the AI industry doesn’t fear regulation, it does fear something else. Naomi Klein and Astra Taylor’s illuminating new bookEnd Times Fascism highlights an “American Dynamism” conference hosted by the venture-capitalist firm Andreessen Horowitz in March. There, Palantir CEO Karp warned that encouraging the expectation that AI would “take away everyone’s white-collar job” was playing with fire. “If you don’t think that’s gonna lead to nationalization of our technology,” Karp said with typical class, “you’re retarded.”
Karp’s comments came at a moment when Anthropic and the Pentagon experienced their first clash. Anthropic sought to set some guidelines around military incorporation of its AI, particularly abjuring mass surveillance and autonomous weaponry. Self-styled War Secretary Pete Hegsethrejected the company’s assertion of any prerogative. Karp’s solution to his vocalized fear of nationalization was to integrate AI aggressively and unapologetically within the militaryand ICE, and upbraided his competitors for any reluctance to serve the national security state. Such reluctance is overstated. After all, Anthropic had already won one of the Pentagon’sfrontier AI contracts, worth up to$200 million each. Karp’s desired military integration builds on a Pentagon trend established a decade ago, when Hegseth’s Obama-era predecessor Ashton Carter headquartered his Defense Innovation Unit “accelerator” in Mountain View. Carter also brought in Google’s Eric Schmidt asan “innovation” advisor.
Still, Hegseth has embraced AI with uncritical enthusiasm. After introducing an AI platform calledGenAI.mil – a version of Google’s Gemini – Hegseth hung posters in the Pentagon reading“I Want YOU To Use AI.” He called it “America’s next Manifest Destiny.” Hegseth debuted his vision of AI integration in the Iran war in March. The use of Palantir’s Maven AI platform for generating targets at scale, aBloomberg investigation recently found, drove the massive bombing of the war’s early weeks. Reliance on a platform that Palantir has for years pitched as capable of finding targets that outperforms both human analysts and rival technology led the military to misidentify outdated satellite imagery of the Minab school as a military installation, according to the investigation. While Palantir disclaimed responsibility, on the garbage-in, garbage-out principle, this mass killing of Iranian children testifies to the tendency of analysts to de facto view the results spat out by Maven as legitimate. Not one to be dissuaded, Hegseth recently appointed Musk and autonomous-drone manufacturer Palmer Luckey to lead a government study project on “identifying the capabilities required to achieve absolute technological dominance on the next-generation battlefield,“ which will inevitably be their own.
But Karp is wrong to think of frictionless integration into the Defense Department as a bulwark against nationalization. Instead it is an accelerant. Hegseth, Karp, Musk, Amodei, Altman, and the rest are ultimately building a military dependent on artificial intelligence. The AI giants and two consecutive Pentagon administrations share the perspective that AI is a resource with geostrategic implications, with the stakes being economic and military dominance belonging either to the U.S. or to China. And whatever degree of kabuki was on display during Hegseth’s brief clash with Amodei, the episode provides a useful and clarifying legacy. It established that there is a fundamental and unresolved sovereignty claim between the U.S. and the AI labs: who controls whom? Who answers to who? How can a state leave itself in the intolerable position of a private entity possessing substantial if not veto power over military decision-making, whatever the merits of those decisions? As long as AI remains privately controlled, this question will persist and only grow in proportion to the Pentagon’s deepening AI dependency.
