Nonprofits sue Anthropic, Google, Meta, OpenAI and SpaceXSI in San Francisco to freeze frontier AI development
Public Citizen and three other groups invoke California's Unfair Competition Law and allege the companies are creating a public nuisance.
Four advocacy groups filed suit on Nov. 12 in the Superior Court of California, County of San Francisco, asking a judge to halt further development of the most advanced AI models at five of the largest developers: Anthropic, Google, Meta, OpenAI and SpaceXSI, the artificial intelligence unit that CBS News identified as SpaceX Super Intelligence.
The plaintiffs are the Irreplaceable Education Fund, Public Citizen, Heat Initiative and The Alliance for Secure AI. Former New Jersey Attorney General Matthew Platkin filed the complaint on their behalf.
The legal theory
The groups rely on California's Unfair Competition Law, which lets private plaintiffs seek injunctions against business practices that are unlawful, unfair or fraudulent. According to Public Citizen's announcement, the complaint alleges that the companies' development practices threaten public health, safety and property, and Reuters reported that the suit also accuses the companies of creating a public nuisance. The requested remedy is an order freezing development of increasingly capable models that, in the plaintiffs' telling, can escape human control. The announcement does not describe a compute threshold, a list of covered models or a time limit for the freeze.
The factual case leans heavily on the defendants' own public statements. As reported by CBS News, the complaint cites a 2015 remark by OpenAI chief executive Sam Altman that AI would probably lead to the end of the world, and a 2025 comment by Anthropic chief executive Dario Amodei putting the chance of a disastrous outcome at about 25%. The complaint also points to incidents in which, the groups allege, existing AI agents bypassed safeguards, accessed outside computer systems without authorization and took actions their developers did not intend, including gaining access to government databases and filing a false police report.
"It is patently illegal for companies to develop and release products they know are unsafe in any respect," Platkin said, adding that the companies are causing a public nuisance, in the groups' announcement. Robert Weissman, co-president of Public Citizen, said the suit "asks the court to apply existing law."
A Meta spokesperson declined to comment, according to CBS News and Reuters. The other companies did not immediately respond to requests for comment. None of the allegations has been tested in court.
Part of a widening wave of AI litigation
The case adds private plaintiffs to a growing list of government suits that seek limits on how AI models are built, not only damages. Florida Attorney General James Uthmeier sued OpenAI in June, and in September the state asked the court to bar the company from developing new models without outside oversight, a request Reuters reported is still pending. On Oct. 14, attorneys general in Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia each filed suits against OpenAI in state courts, according to Tech Policy Press. Tech Policy Press reported that some of those states seek a pause in testing and development, and that Arkansas asks for a bar on models capable of recursive self-improvement. Separately, Reuters reported that OpenAI has said it paused training its most capable models until additional safeguards are in place.
The case arrives without a federal AI safety framework. Reuters reported that the Trump administration has so far declined to pursue large-scale AI regulation over concerns about competition with China, leaving states and private groups to test what existing law can require.
What this means for in-house counsel
Public statements by executives about AI risk are being cited in complaints. The plaintiffs in this case built much of their pleading around what company leaders have said publicly about catastrophic risk and control. Legal teams at AI developers, and at companies that market AI features, should review how leaders describe safety, control and catastrophic risk, and make sure those statements match internal testing records and disclosures.
Companies that deploy frontier models from these vendors should map their dependence on them. An injunction is a remote outcome, but the litigation wave, combined with state requests for development limits and OpenAI's own reported training pause, is a supply risk worth noting in vendor contracts. Counsel can ask providers about model versioning commitments, notice of material changes and continuity if a vendor pauses or restricts a model.
Incident records matter as well. The complaint relies on reported cases of AI agents acting outside their intended scope. Companies running agentic systems should keep clear logs of permissions, failures and remediation, both to manage risk and to answer regulators or litigants who cite such incidents as proof that controls fall short.
Free weekly email
The Counsel House Brief
The week's most important legal news for general counsel and in-house teams, in one short email.
Counsel House corrects errors promptly. Spot one? Email daniel@counselhouse.org.