Four AI giants just got sued for slowing down

On September 18, 2026 (UTC-7), four consumers filed a class-action lawsuit in the U.S. District Court for the Northern District of California accusing Anthropic, OpenAI, SpaceXAI, and Google of violating Section 1 of the Sherman Act. Their theory: the companies publicly coordinated to slow the pace at which their competing frontier AI products improve, and paying subscribers are being shortchanged because the models they paid for aren't getting better as fast as they could.

Consumers suing major AI companies over alleged slowdown collusion

The complaint

The case, Buist v. Anthropic PBC (3:26-cv-10693), names four defendants: Anthropic PBC, OpenAI OpCo LLC, SpaceXAI LLC, and Google LLC. The plaintiffs — Charles Buist and Nick Spetsas of Florida, and Cheyenne Hunt and Christine Bullock of California — each bought paid subscriptions to ChatGPT, Claude, Grok, and Gemini (Bullock only Claude).

The core legal theory is straightforward. When competitors agree to reduce the quality of their products or the rate at which those products improve, that's an agreement to restrict output — the kind of restraint Sherman Act Section 1 was written to punish. The plaintiffs plead it as unlawful per se, and in the alternative under quick-look and rule-of-reason analysis. They're seeking treble damages under the Clayton Act and an injunction barring any horizontal agreement about training compute, release delays, capability checkpoints, or exchanges of competitively sensitive information.

The market definition is telling: paid consumer subscriptions to general-purpose frontier AI assistants. The four defendants collectively account for at least 80% of that market, the complaint alleges. ChatGPT Plus runs about $20 per month.

The timeline that led here

This didn't start with a random tweet. The complaint traces a deliberate arc:

Why this lawsuit matters

The plaintiffs are careful to say they take AI safety seriously. They explicitly do not challenge any defendant's unilateral safety decisions, testing, environmental impact, or its own development pace. They don't challenge lobbying Congress or the White House. Their line is specific: safety guardrails should be set by the public through regulation and juries, not by four companies that collectively control 80% of the market deciding among themselves how fast their products improve.

That's a sharp distinction, and it's the one the defendants will have to navigate. Amodei's essay itself conceded that the second step — coordination among frontier companies to establish common safety standards and limits on the rate of progress — "depends on government mediation or antitrust waivers." The footnote acknowledged the legal risk. Altman then said OpenAI would start working anyway, without waiting for the exemption. That sequence — public coordination, acknowledgment that antitrust might apply, and then proceeding anyway — is exactly what the plaintiffs say constitutes a conspiracy.

The market definition also matters. The complaint isn't about enterprise contracts or research labs. It's about the $20/month ChatGPT Plus subscriber who was promised a product that keeps getting better and now has four companies saying, in public, that they're going to slow it down. Whether that's a legally cognizable antitrust injury is the question.

This lawsuit lands in a genuinely awkward spot for the industry. The companies didn't secretly collude in a smoke-filled room. They said all of this out loud, on purpose, because they wanted the safety movement to look like a coordinated industry position. The problem is that antitrust law doesn't care whether your motives are noble — it cares whether competitors agreed to restrict output. A CEO can unilaterally decide to slow down his own company. Four CEOs agreeing to slow down together is a different legal question.

The "we're doing this for safety" defense is not as strong as the companies might hope. Price-fixing defendants don't get a pass because they argued higher prices were good for consumers. The Sherman Act is blunt about horizontal agreements among competitors, regardless of intent. The plaintiffs' per se theory is aggressive, but it's not frivolous — public coordination to reduce the rate of product improvement is the kind of thing courts take seriously.

Expect the defendants to move quickly to dismiss, arguing that public statements about safety don't constitute an agreement, that the market definition is too narrow, and that the plaintiffs can't show they were harmed. But even getting to discovery will be ugly. The emails, meeting notes, and working group records from July onward are exactly what plaintiffs' lawyers will want.

What to watch