The Slowdown Cartel

On Friday, 18 September 2026, four people who pay for ChatGPT, Claude, Grok or Gemini sued the companies that make them. The suit, filed in the U.S. District Court for the Northern District of California as No. 3:26-cv-10693, a Clayton Act antitrust case with a jury demand, on behalf of a proposed nationwide class of paid subscribers, names Anthropic, OpenAI, SpaceXAI and Google. The charge is that they agreed to slow down.

It is a price-fixing theory with the price replaced by progress. An agreement among the chief rivals that their progress “should be slower than competition would otherwise produce has an anticompetitive effect on consumers,” the plaintiffs argue. The subscribers pay for a product. The product was allegedly agreed to improve more slowly.

The meeting of the minds, on X

Cartels used to meet in hotel rooms. This one, by the complaint’s account, met in public.

The coordination “largely took place on Sept. 12,” the lawsuit argues, when Dario Amodei published “We Must Pace the Frontier”, and Sam Altman, Elon Musk and Google DeepMind’s Demis Hassabis each publicly agreed the same day. The posts are short. Altman: “I agree with Dario that we need to pace the frontier.” Musk: “Dario is right.”

The complaint reaches back further, to a July 2026 statement signed by senior employees of several labs that acknowledged “intense competitive pressure not to unilaterally slow” and asked government to back a global effort to slow automated AI development.

The essay saw the problem coming. “For antitrust reasons,” Amodei wrote, “it’s helpful for the US government to mediate or at least enable these discussions — they don’t need to participate, but do need to issue a narrow waiver for certain kinds of safety conversations.” Altman declined to wait for one. OpenAI welcomed a “federal framework that sets consistent safety requirements,” he said, but “we do not believe we need to wait for an antitrust exemption or legislation to begin the work of providing this confidence.”

One defendant asked for the waiver. Another said it would go ahead without one. Six days later they were co-defendants.

The plaintiffs are not against safety

That is the part that makes the case worth reading. The plaintiffs do not object to any company deciding on its own to slow down. They do not object to the companies asking Congress or the White House to regulate AI, or asking for an antitrust exemption. What they say antitrust law forbids is the “shortcut” of agreeing to “substitute collective restraint for individual accountability.”

Their lead attorney, Nick Rowley, makes the case for danger more bluntly than the defendants do: “AI will quickly spin out of human control and could kill us all if we allow AI safety and protocol … to be controlled by private self-serving agreements between the world’s most powerful ‘for profit’ technology companies.”

Read as a document, the complaint carries both arguments that have been fighting all year over AI policy. The technology could kill us all. The people who say so should not be the ones who decide, together, how fast it goes.

The labs’ case, in their words

Amodei’s essay makes the defendants’ case better than any brief will. “Not building the technology deprives humanity of benefits or simply places AI in the hands of authoritarian powers, while building it too fast is reckless,” he wrote. “To be clear, pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this.” And he named the solution that would make the lawsuit moot: “The most effective method of pacing is via regulation that targets all US frontier AI companies, as that covers even those who are unwilling to cooperate voluntarily.”

Altman, at the Security Council five days after the suit, argued the race was a choice, not a law: “Beating companies in competitive race is not a reason to make rash decisions, nor do we believe we are locked in a race where we are unable to do that.”

A statute binds everyone and needs no agreement. A pledge among rivals binds only the rivals, and an agreement among rivals is what the Sherman Act exists to read.

The smaller firms’ case

Smaller firms made the plaintiffs’ economic argument within the week. To NPR, Cohere co-founder Nick Frosst said “a blanket, privately coordinated slowdown is not a neutral safety policy, it would lock in the advantage of the small number of labs that already have the most compute, capital, and distribution.” David Bellamy of the UAE-backed Institute of Foundation Models, to the same reporter: “whoever has the strongest model today is going to dominate the market share.” Former FTC Commissioner Alvaro Bedoya: “When powerful incumbents lock up the market, the rivals, the upstarts, the scrappy players in that market, can no longer compete on product or quality and service. They are just locked out.”

The White House’s case, from the other direction

The administration reached the same suspicion by another road. Two days after the essay the President asked: “Concerning AI, when, in the History of Business, did anyone see the Leaders of an Industry call for Regulation that, if strongly implemented, will drive them into oblivion and bankruptcy?” The Vice President called it “a bit of a Trojan horse.” David Sacks, co-chair of the President’s science council, reportedly told the executives to “stop pretending you need anyone else’s permission” to slow down.

The plaintiffs’ lawyer, a former Biden-era FTC commissioner, a Canadian startup and the Trump White House all arrived at the same sentence: when the leaders agree to go slower, the leaders stay the leaders.

The defendant who promised to win

One defendant has made the cartel theory harder to sell. A week after the suit was filed, SpaceXAI’s owner said his company would “reach pole position in about 6 months” and ship a model equivalent to Anthropic’s Fable or OpenAI’s GPT-6 in two to three months. Three days after that he called a chart ranking his Grok models three tiers below theirs “accurate,” with the caveat “for now.”

An alleged member of a slowdown pact announced, in public, that he intended to pass the other members.

Eleven days later

On 29 September, three of the four defendants’ chiefs, Amodei, Musk and Google’s Sundar Pichai, and OpenAI’s president Greg Brockman, sat down at the White House and signed a joint commitment on frontier safety, hosted by the government. One sentence of it, in the full text as reprinted by Forbes Australia, reads: “The participating companies will meet regularly to establish standards and best practices to improve the safety of their systems.” Whether that is the “mediate or at least enable” Amodei asked for is a question for the court. What the record shows is where it was signed.

The labs asked for a federal framework. The government declined to write one, then hosted the meeting where they signed a joint commitment anyway. The only party that has asked a judge to rule on any of it is the customer.

The receipts (free, on this site): Dario Amodei · Sam Altman · the funding ratchet