AI slowdown lawsuit turns safety coordination into an antitrust fight

A new lawsuit has turned the AI industry’s most sensitive safety debate into an antitrust fight. AP reported that a proposed class action filed in the U.S. District Court for the Northern District of California accuses Anthropic, OpenAI, SpaceXAI and Google of making an illegal agreement to slow the pace of artificial intelligence development. The plaintiffs are paid subscribers to ChatGPT, Claude, Grok or Gemini who argue that a coordinated slowdown would reduce the value of services they buy because competition would otherwise push the companies to improve their products faster.

The case centers on the public response to a September 12 essay by Anthropic CEO Dario Amodei, who called for industrywide cooperation to pace frontier AI progress and strengthen safety measures. According to AP, the lawsuit says OpenAI’s Sam Altman, SpaceXAI’s Elon Musk and Google DeepMind’s Demis Hassabis each publicly supported the idea that day. The plaintiffs argue that agreement among direct rivals to slow development has an anticompetitive effect, even if the stated motivation is safety.

The filing has not proven wrongdoing, and the companies had not immediately responded to AP’s request for comment. That distinction matters because safety coordination is not automatically illegal. Companies can publish their own policies, support standards, fund audits or decide individually to delay a release. The legal question is whether rivals crossed a line from independent safety choices into an agreement that restrained competition in paid consumer AI services.

The case highlights a problem policymakers have not yet resolved. AI labs say the most capable systems could create risks that no single company can manage alone, especially if competitive pressure rewards speed over caution. At the same time, antitrust law is designed to prevent competitors from coordinating behavior that harms consumers. Amodei had acknowledged in his earlier proposal that government mediation or a narrow waiver could help safety conversations avoid antitrust concerns. Without a public framework, however, even a safety pact can be framed as a cartel by opponents.

For consumers, the complaint makes an unusual argument about product value. Paid AI subscriptions are sold partly on access to better models, new tools and frequent capability upgrades. If major labs agreed to slow those improvements, plaintiffs say subscribers would be paying for less than a competitive market would have delivered. The companies are likely to argue that safety work, evaluation and controlled releases also improve value by making systems more reliable and less likely to cause harm.

The broader stakes go beyond damages for subscribers. If the lawsuit advances, it could shape how AI companies discuss joint standards, outside audits, compute limits and release checkpoints. A ruling against coordination could make labs more cautious about public safety commitments. A ruling that leaves space for supervised cooperation could push Congress or regulators to create clearer rules for when competitors may discuss AI risk. Either way, the case shows that AI governance is now being contested not only in policy forums and research labs, but also in competition courts where consumer harm, market power and safety claims will be tested against one another.