← All stories

Law and policy

Slowing AI down for safety is an illegal cartel, a new lawsuit claims

On September 18 four paying subscribers filed a proposed class action accusing Anthropic, OpenAI, SpaceXAI and Google of breaking US competition law by backing a plan to pace AI development.

By · News from Sep 18, 2026 · 4 min read · Checked Oct 2, 2026

Here is a puzzle worthy of a law exam. In September, the heads of several big AI companies publicly agreed that the most powerful AI should be developed more slowly, for safety's sake. Six days after that agreement began, four of their customers sued them, arguing that rivals agreeing to slow down is itself against the law.

What the lawsuit says

On September 18, 2026, four paying subscribers to AI services, two from Florida and two from California, filed a proposed class action. That is a lawsuit brought on behalf of a whole group of people with the same complaint, here paid subscribers. They filed it in the federal court for the Northern District of California, and the case is called Buist v. Anthropic PBC (case number 3:26-cv-10693). The defendants are Anthropic (maker of Claude), OpenAI (maker of ChatGPT), SpaceXAI (Elon Musk's AI business, which makes Grok) and Google.

The complaint traces its case to Anthropic chief executive Dario Amodei's September 12 essay, which called for the industry to pace the frontier (meaning slow the progress of the most advanced AI models), and to the public support that followed from OpenAI's Sam Altman, Musk and Google DeepMind's Demis Hassabis. In the plaintiffs' view, those public endorsements add up to an illegal deal. According to Yahoo Finance, the complaint calls it "a classic output-restricting cartel."

Antitrust, in plain English

Antitrust law, called competition law in many countries, exists to stop companies from teaming up against their customers. This suit relies on the Sherman Act, a US law from 1890 that bans agreements between competitors that hold back fair trade.

A cartel is a group of competitors who coordinate instead of competing. The textbook example is rivals agreeing to fix prices. An output-restricting cartel is one where competitors agree to produce less, which usually means customers get less, or pay more, than they would in an open market. The plaintiffs argue that a coordinated slowdown works the same way: rival companies jointly holding back better products.

Safety versus competition

The irony is the whole story. Supporters of the slowdown frame it as a safety measure. The lawsuit says that, whatever the motive, rivals coordinating to hold back their products is exactly what competition law was written to stop. Both sides are pointing at the same public statements and reading them in opposite ways.

Two details are worth noticing. Microsoft chief executive Satya Nadella had also welcomed a slower pace, while warning that any such system should not be run by a small handful of players. Microsoft is not among the defendants named in the reports. And a court will eventually have to tackle a basic question: is publicly agreeing with a rival's idea the same as making an agreement with that rival? Nothing has been decided yet.

For now, the case is brand new. It was covered by the legal news services Bloomberg Law and Law360, but as of October 2 there had been no rulings, and we found no public response from any of the four companies.

What we don't know yet

Sources

Keep reading

Law and policy

The Justice Department backs OpenAI on fair use; weeks later an appeals court rejects fair use in a narrower AI case

On September 1 the Justice Department told a court that training AI on copyrighted work is fair use, surprising two other federal agencies; on September 29 an appeals court rejected fair use in a case about a non-generative legal-research tool.

Sep 29, 2026 · 4 min read

New to AI? Start with which assistant to use, how to stop AI training on your chats, which ChatGPT plan to pay for or how to write a good prompt.