An alliance of four nonprofit organizations sued five US AI developers — Anthropic, Google, Meta, OpenAI and SpaceX's AI business — in San Francisco Superior Court on November 12, asking the court to stop them from building more advanced AI models. According to Guancha.cn's November 13 report, the plaintiffs accuse the developers of racing ahead with frontier AI despite knowing it may pose catastrophic risks, and of starting to let AI take part in its own development.

Courtroom and AI technology elements, illustration
The lawsuit pushes the frontier-model safety debate into the courts

The Alliance for Secure AI, the Irreplaceable Education Fund, Public Citizen and the Heat Initiative filed jointly. Alliance CEO Brendan Steinhauser said in a statement that executives had spent years warning about catastrophic risk from their own frontier models while knowing the systems "do not always act as developers intend" and that no reliable method exists to control them. The complaint also argues that as companies lean on current models to build the next generation, the risk of "recursive self-improvement" grows.

The filing leans heavily on the defendants' own words: Sam Altman said in 2015 that AI could "quite possibly lead to the end of the world," and Dario Amodei said in 2025 that AI development carries a 25 percent chance of catastrophe. In September, Amodei launched an initiative calling on the industry to slow frontier performance gains, with Altman and Musk voicing support.

Development and dealmaking have hardly slowed. Anthropic is still pursuing an IPO at a target valuation of up to 1.4 trillion dollars, expected to list on Nasdaq on November 19, after releasing Claude Fable 5.2 and Mythos 5.2 on November 9 and signing a six-year, 13.7-billion-dollar compute contract with the RUM Group. Musk said on X on November 10 that Grok 4.8, with 2.5 trillion parameters, had finished pretraining and entered reinforcement learning. OpenAI is working with Australia's Firmus on dedicated compute at a Malaysia AI factory.

It is not the first courtroom fight. Florida's attorney general sued OpenAI in June over ChatGPT safety, and eight more states filed suits in October. But where earlier cases centered on concrete harm, this one asks a court to halt frontier development itself — moving the tug-of-war between the AI race and risk control squarely into the judicial arena.