AI Labs Face an Antitrust Suit—A Filed Complaint Is Not Proof of Collusion

Four individuals filed a federal antitrust complaint against Anthropic, OpenAI OpCo, SpaceXAI and Google on September 18, 2026, according to the Northern District of California docket. It identifies Charles Buist, Cheyenne Hunt, Christine Bullock and Nick Spetsas as the plaintiffs and lists the cause of action as litigation under Section 1 of the Sherman Act.

The plaintiffs allege that the companies agreed to slow improvements in their competing AI services, diminishing the value of paid subscriptions. What has been established is narrower: a complaint has been filed and a case has opened. The filing is not a judicial finding that an agreement existed, competition was restrained or subscribers suffered compensable harm.

What the subscribers allege

The proposed class action treats public discussion of slowing advanced AI development for safety reasons as evidence of coordination. The Associated Press account published September 19 says the plaintiffs subscribe to ChatGPT, Claude, Grok or Gemini and links the alleged coordination to public responses made on September 12 to an essay by Anthropic CEO Dario Amodei.

The complaint characterizes responses from OpenAI CEO Sam Altman, SpaceXAI CEO Elon Musk and Google DeepMind co-founder and chair Demis Hassabis as assent to industry-wide coordination. Those statements are evidence selected and interpreted by the plaintiffs, not findings accepted by a court. The central factual dispute is whether the executives expressed compatible policy views or formed an understanding about how quickly competing products would improve.

As for the defendants’ position, the same AP account states that representatives of Anthropic, OpenAI, Google and SpaceXAI had not responded to its requests for comment by Saturday. That silence is neither an admission nor a substitute for the answers or motions the companies may later place on the court record.

Why safety discussions can create an antitrust question

Section 1 addresses concerted conduct rather than decisions made independently by separate businesses. The Justice Department’s overview of federal antitrust law explains that the Sherman Act prohibits agreements that unreasonably restrain trade and identifies price fixing, bid rigging and market allocation as examples of agreements among competitors that can constitute criminal violations.

This lawsuit advances a different, civil theory: the plaintiffs characterize a supposed agreement to slow product improvement as a restriction on competitive output. Their claimed harm follows from that theory. A provider that slows independently risks losing subscribers to faster-moving rivals; if rivals collectively limit improvement, the complaint argues, customers lose that competitive alternative while continuing to pay.

Safety cooperation does not resolve the legal question in either direction. Competitors can advocate regulation, discuss public risks and independently adopt similar precautions without necessarily reaching a commercial agreement. But an actual understanding that replaces independent product decisions with collective restraint can raise antitrust concerns even if participants describe their objective as safety.

Claim versus record

• Parties: The case names four individual plaintiffs and four defendants: Anthropic, OpenAI OpCo, SpaceXAI and Google. Being named as a defendant does not establish liability.
• Statute: The complaint invokes Section 1 of the Sherman Act. The plaintiffs must still establish concerted conduct and the other elements required by their theory.
• Requested remedies: Bloomberg Law’s coverage of the complaint says the plaintiffs seek class certification, an injunction and a declaration that the alleged conduct violated federal antitrust law. None of those requested remedies is a finding on the merits.
• Alleged coordination: The plaintiffs rely on public endorsements of Amodei’s proposal and related safety discussions. No cited court ruling has accepted their inference that those statements created an agreement.
• Defendants’ response: No substantive response from any defendant was included in the initial public coverage. Formal court responses remain to be filed.
• Established result: A proposed class action exists. The public record cited here does not establish class certification, an injunction or an adjudicated antitrust violation.

What remains unproved

The complaint does not establish that any defendant actually slowed a product, that the companies adopted matching development schedules or that public expressions of support produced a binding commitment. It also does not determine the amount of any subscriber loss or whether the proposed nationwide class can satisfy the requirements for certification.

Questions about the existence of an agreement, the meaning of the cited statements, the relevant market, causation, standing and class treatment remain open. They are potential issues for litigation, not positions that should be attributed to the defendants before those companies make them in court.

The procedural distinction is decisive. Similar language or conduct can arise from independent responses to the same safety concerns; the plaintiffs’ Section 1 theory depends on proving concerted restraint rather than parallel policy preferences. The next meaningful public developments will be proof of service and the defendants’ formal responses, followed by any ruling on whether the complaint may proceed.

For now, the accurate description is consequential but limited: four consumers have brought a federal antitrust case accusing major AI providers of coordinating a slowdown. The September 18 filing created a dispute for a court to examine; it did not convert public alignment on AI safety into proven collusion.

Also read:

• Enterprise AI’s 8.3× Usage Gap Is a Proxy—not Proof of Value
• Claude Closed 65% of a Safety Gap in 60 Hours—but Not the Last Mile
• Keep Your Character: Removing Moles and Spots Without Losing Your Face