Subscribers sue four AI labs, calling their pacing pledge a cartel

Four paying subscribers to the defendants’ AI products filed a proposed class action, Buist v. Anthropic, PBC, against Anthropic, OpenAI, Google and xAI in the US District Court for the Northern District of California (No. 3:26-cv-10693), alleging that the companies’ public agreement to “pace the frontier” amounts to an unlawful conspiracy to restrict output under Section 1 of the Sherman Act. The complaint is pleaded as per se illegal and, in the alternative, under quick-look and rule-of-reason analysis; it seeks treble damages under the Clayton Act, a declaratory judgment, an injunction against the labs coordinating on development pace, training compute, release timing or capability checkpoints, and a jury trial.

The complaint’s theory treats coordinated slowdown as no different from coordinated price-fixing: an agreement among rivals about how fast their competing products improve, it argues, restricts the supply of a good — model capability — for which no antitrust exemption exists. It builds its chain of evidence from public statements rather than leaked documents: a July working group among labs on shared standards, Google DeepMind chief executive Demis Hassabis’s proposal for an industry standards body, OpenAI chief scientist Jakub Pachocki’s September essay naming coordinated slowdown as an option, a Wired report that OpenAI had sought Congressional guidance on the antitrust implications of pacing, and, centrally, Anthropic chief executive Dario Amodei’s 12 September essay calling for labs to slow capability gains, which OpenAI’s Sam Altman and xAI’s Elon Musk publicly endorsed within hours. The complaint also cites OpenAI communications adviser Joe Lehane’s acknowledgement, three days after Amodei’s essay, that “weeks of talks” had taken place between the companies.

None of the defendants had responded in court filings at the time of the suit. The case tests whether a safety commitment voiced as competitors agreeing, in public, to slow each other down can itself become the antitrust violation the industry’s critics have otherwise accused the labs of avoiding — an inversion of the usual complaint that AI companies race too fast rather than coordinate to go slow.