AI Slowdown Lawsuit Accuses Four Labs of Collusion
3 min readFour paying AI subscribers have sued Anthropic, OpenAI, Google and SpaceXAI, arguing that the industry’s recent public calls for a safety slowdown were something else entirely: an illegal agreement among competitors to hold back the products those customers already pay for. The complaint was filed on September 18 in the US District Court for the Northern District of California.
How the slowdown talk started
The proposed class period opens on September 12, the day Anthropic chief executive Dario Amodei published an essay urging the industry to cooperate on decelerating capability gains in favor of stronger safety work. Rival labs weighed in over the following days, and the idea of a coordinated pause moved quickly from a fringe position to a mainstream talking point.
That conversation followed a run of incidents in which AI agents behaved in ways their developers had not anticipated. Safety researchers welcomed the change in tone. Antitrust lawyers, it turns out, were reading the same essays.
What the lawsuit alleges
The plaintiffs argue the four companies coordinated how quickly their competing products would improve, and that subscribers kept paying unchanged monthly prices for a rate of progress the defendants had agreed to throttle. Under US antitrust law, an agreement among rivals to restrict output or quality is treated as unlawful on its face, whatever the stated motive.
The proposed class covers anyone in the United States who bought a paid individual consumer subscription directly from a defendant for ChatGPT, Claude, Grok or Gemini from September 12 onward. According to CBS News, the plaintiffs want a court order barring the companies from coordinating limits on training, release schedules or other development decisions, while leaving each firm free to adopt its own safety measures and comply with regulators. They are also seeking damages that can be tripled under federal law, plus attorneys’ fees.
Why the AI slowdown lawsuit matters
This is the first real test of a question the industry has circled for a year: can competing labs talk to each other about pacing without that conversation becoming a cartel. Safety coordination and price fixing can look uncomfortably similar from the outside, and courts have historically been unsympathetic to defendants who argue that restraining competition was good for the public.
The practical risk is chilling. If a plausible antitrust theory attaches to any cross-lab safety discussion, the labs most inclined to talk are the ones most exposed. Watch for whether the defendants move to dismiss on First Amendment grounds, and whether any regulator opens a parallel inquiry.
The case is early, and an allegation is not a finding. But the filing has already converted a philosophical debate about how fast AI should move into a question a federal judge will have to answer.