The class-action complaint was filed Friday in the U.S. District Court for the Northern District of California: Anthropic, OpenAI, SpaceXAI and Google stand accused of antitrust collusion - not for fixing prices, but for fixing the pace. The theory, per The Hill: when Dario Amodei published his September 12 essay calling on the industry to "pace the frontier," and Sam Altman, Elon Musk and Demis Hassabis endorsed it within hours, four competitors made an agreement in public. Under the Sherman Act, the plaintiffs argue, agreeing to slow improvement is as unlawful as agreeing to split a market.
The essay and the endorsements are the evidence. Amodei wrote "we must slow the space in which we improve the capabilities of AI models." Musk replied within hours that "Dario is right." Altman said he agreed. Hassabis called it the right path forward. The complaint reads those four statements as the outline of a concert of action among the four labs that control nearly all frontier capability - an agreement that, if it holds, would raise the price and lower the quality of the tools their customers already pay for.
The plaintiffs are four individuals, two of them Florida attorneys, seeking class status for paid subscribers of Claude, ChatGPT, Grok and Gemini. Their damages theory is almost poetic: customers paid for tools that would have gotten better faster, and did not. Attorney Nick Rowley told Politico the point is that safeguard decisions belong to government, not to "private self-serving agreements between the world's most powerful for-profit technology companies."
The irony stack is now three layers deep. The lab that asked for brakes is the lab with four incident disclosures this year. The essay that asked for coordination came with a request for a narrow antitrust waiver for safety conversations - which Senator Josh Hawley rejected in a Tuesday hearing with the words "absolutely no way that's happening." And one week after the pact formed in public, the pact itself is a federal antitrust defendant.
What happens next is the test of the whole arc this paper has tracked since No. 73. If the suit survives a motion to dismiss, every future safety coordination among frontier labs will be drafted by lawyers first and scientists second. If it dies, the labs will have learned that public promises cost nothing. Either way, the era in which lab chiefs could coordinate the frontier's speed in plain prose is probably over - which may be exactly what everyone involved secretly wanted, and exactly what none of them can say.
The Waiver Problem
The defense will be built on the waiver. Amodei's essay asked Congress for a narrow safe harbor - protection for safety conversations among competitors - and Hawley's hearing answered no. That sequence now reads like the plot of the complaint: the labs asked for permission to coordinate, were refused, and are being sued for the coordination they had already begun in the open. What happens to the waiver request is now inseparable from what happens to the suit. A safe harbor granted mid-litigation would be Congress picking a winner. A refusal leaves every future safety pact with the structure of a conspiracy - a problem the labs wrote for themselves in public.
The Takeaways
- A class-action antitrust suit filed in the Northern District of California accuses Anthropic, OpenAI, SpaceXAI and Google of colluding to slow AI capability gains.
- The evidence is the public record: Amodei's pace-the-frontier essay and the same-day endorsements from Altman, Musk and Hassabis.
- Plaintiffs are four subscribers seeking class status - customers who paid for tools that would have improved faster absent the pact.
- The waiver request that would have legalized safety coordination was refused days earlier - every future pact now drafts around the Sherman Act.
