Four Labs Are in Court Over Agreeing to Slow AI Down, and the Complaint Leans on That Antitrust Waiver Request

Four paying subscribers — Charles Buist, Cheyenne Hunt, Christine Bullock and Nick Spetsas — filed a proposed class action on September 18 in the U.S. District Court for the Northern District of California against OpenAI, Anthropic, Google DeepMind and SpaceXAI, before U.S. Magistrate Judge Nathanael M. Cousins. The suit targets the September 12 essay by Anthropic CEO Dario Amodei proposing to "pace the frontier," whose three-step plan included placing outside evaluators inside AI companies, reaching industry-wide agreement on shared safety benchmarks, and setting constraints on how fast capabilities advance; that same day OpenAI's Sam Altman, SpaceXAI's Elon Musk and Google DeepMind's Demis Hassabis each publicly voiced agreement. The complaint also alleges the coordination took shape months earlier, citing a July 2026 statement signed by high-ranking employees at several leading labs acknowledging "intense competitive pressure." The legal basis is Section 1 of the Sherman Act — an agreement among competitors that unlawfully restricts competition. The plaintiffs draw an explicit boundary: they do not object to any single company deciding on its own to slow down for safety, but argue antitrust law does not permit the shortcut of substituting collective restraint for individual accountability, with the complaint stating that the antitrust laws do not permit competitors to decide among themselves that competition is too dangerous. They say subscription prices stayed the same while capability gains slowed, and ask the court to bar coordination on training limits, product releases or other development decisions while preserving each company's ability to adopt its own safety measures and comply with government requirements, plus damages that can be tripled under federal antitrust law and attorneys' fees. The defendants have not filed substantive responses.

A Sentence Written Four Days Ago Now Runs Through the Complaint's Logic

Writing on September 19 about the three labs discussing a self-run pre-release testing body, this site argued that the most informative item was not any objection but a request from the proposing side: alongside pushing for aligned benchmarks, Amodei had sought antitrust protection for safety-related coordination. The judgment then was that **this amounts to conceding the coordination has competition-restricting effects — otherwise no exemption would be needed** — followed by the line that it needs an adjudicator not defined by those three. Four days later an adjudicator arrived, and it is not the government. It is private litigation. The complaint's reasoning nearly overlaps that passage: if even the proposing side believes an exemption is required, then beginning to coordinate without one is itself the problem. Altman's reply at the time now cuts both ways. What he said was that they welcome a federal framework setting consistent safety requirements for frontier AI, but do not believe they need to wait for an antitrust exemption or legislation to begin the work. The intent was to convey resolve; placed in the complaint's context, it reads like an acknowledgment that the work began without an exemption.

The Plaintiffs Are Not Opposing Slowing Down, and That Is the First Thing Lost in Retelling

"Lawsuit opposes slowing AI down" invites the reading that this is the accelerationists striking back at the safety camp. The complaint does not say that. The plaintiffs state explicitly that they do not object to any company deciding on its own to slow down for safety, and their requested relief preserves each company's ability to adopt its own safety measures and follow government requirements. What they object to is the shortcut: **substituting collective restraint for individual accountability.** So the axis of this dispute is not safety versus acceleration. It is who gets to decide. One company braking bears its own consequences and the market renders a verdict; four companies agreeing to brake together converts a question competition would settle into a question an agreement settles — and the parties to that agreement happen to be essentially all of the supply. This is the legal version of what Cohere's Gomez said in the September 19 story.

This Case Will Turn on Whether Public Essays Count as an Agreement

The legal crux has to be stated plainly, or readers will conclude that publishing an essay is itself unlawful. Section 1 of the Sherman Act requires an **agreement**. What the plaintiffs hold is a public essay dated September 12, three executives' separate public endorsements the same day, and a July statement acknowledging competitive pressure. Whether that constitutes an agreement is the central contested issue — because antitrust law draws a long-standing distinction: **concerted action can be unlawful, while parallel conduct generally is not.** Several companies independently reaching similar conclusions and publicly stating similar positions is, legally, a different thing from privately settling on a shared limit. The plaintiffs' advantage is that the private working groups reported in the September 19 story have been meeting since July; if the case reaches discovery and those materials surface, the line between "said it publicly" and "agreed to it privately" gets concrete. The defendants' advantage is that everything visible today is still each party speaking for itself. The shape of the damages claim deserves its own look: prices unchanged, capability gains slowed — which treats the *rate of product improvement* as a dimension of competition on which to claim injury. That angle is unusual and hard to quantify, and whether it holds up will directly determine whether the case clears class certification.

Whatever the Ruling, One Consequence Has Already Landed

The defendants have not answered substantively and a result is a long way off. But one thing changed as of September 18: **the path of industry self-regulation now carries litigation risk.** On September 18 this site covered three labs walking over to cosign AEF-1, on September 19 their discussions about a self-run pre-release testing body, and on September 20 Anthropic bringing evaluators inside its own company. All three presuppose that the industry can sit down together and write rules. Every step now happens under this shadow — with the notable exception of the third, because bringing evaluators in-house is unilateral. It requires agreeing on nothing with a competitor, which makes it the legally cleanest of these paths. The next node worth watching is the defendants' answer. Will they argue this was parallel conduct, or argue the coordination is justified (and therefore needs an exemption)? Those two defenses lead to entirely different places, and the choice will answer the question the September 19 story left open: what do these companies actually think they are doing.

via: CBS News, PBS News, Law Commentary's review of the filing, Tom's Hardware