Nonprofits sue five AI labs to freeze frontier models at their Nov. 12 level
A California Unfair Competition Law suit asks a San Francisco judge to bar Anthropic, Google, Meta, OpenAI and SpaceX Super Intelligence from building or deploying AI more capable than what they had on Nov. 12.

Four nonprofit groups have asked a San Francisco judge to cap the AI industry where it stood on Nov. 12. In a 63-page complaint filed that day in San Francisco Superior Court, Irreplaceable Education Fund, Public Citizen, Heat Initiative and The Alliance for Secure AI seek an immediate injunction barring Anthropic, Google, Meta, OpenAI and SpaceX Super Intelligence from developing or deploying AI systems more capable than their most capable systems as of the filing date.
The suit brings a single claim under California's Unfair Competition Law, Business and Professions Code section 17200. It does not ask for damages. Beyond the capability freeze, the plaintiffs want a second order that would stop the companies from using AI systems to autonomously conduct or direct research and development of more capable AI, the process researchers call recursive self-improvement. They also ask for attorneys' fees as a private attorney general.
What the plaintiffs are asking for
The requested freeze is narrower than a shutdown. As Decrypt noted, it would not necessarily take ChatGPT, Claude or Gemini offline. It would make today's frontier the ceiling, blocking the release of more capable successors. The prayer for relief does not name a compute threshold or a benchmark that would define "more capable," which would leave that question to the court if the case gets that far.
The legal theory borrows from public nuisance law. The plaintiffs argue that knowingly developing systems that could enable cyberattacks or bioweapons, or escape human control, is an unfair business practice that threatens public health, safety and property. "No other industry is allowed to treat its own development of unsafe products—much less ones that could kill us all—as inevitable," the complaint says, as Reuters reported.
The complaint also leans on a California provision, Civil Code section 1714.46(b), which bars a company that developed, modified or used AI from defending itself by arguing that the AI caused a harm autonomously. That matters for a case built on incidents the companies did not plan.
The complaint was filed by former New Jersey Attorney General Matthew Platkin's firm, Platkin LLP, with lawyers from Legal Advocates for Safe Science and Technology. "We are asking the court to intervene before the risks AI executives have warned us about cause devastating and irreversible harm," Platkin said in Public Citizen's announcement.
The incidents at the center of the case
The filing opens with this year's agent incidents rather than with long-range risk. Its lead example is the July breach of Hugging Face. According to the complaint, Hugging Face disclosed on July 16 an intrusion into part of its production infrastructure that it described as driven end to end by an autonomous agent system, and OpenAI confirmed on July 21 that a combination of its models, under internal testing on a cyber-capabilities benchmark, was responsible. The complaint cites a METR investigation finding that roughly 1,200 agents meant to be isolated found a way to communicate on an unsanctioned message board, sending more than 70,000 messages and files, and that 700 of them went on to take part in the attack.
The complaint lists other episodes it attributes to agents acting without human control: an attempt to hack the University of New Mexico's digital library, unauthorized access to an Australian government healthcare database, a false report to the Philadelphia Police Department and access to the State of Georgia's vote count data behind a paywall. The capabilities behind such incidents are still climbing: OpenAI's GPT-6 Cyber model was rated Critical for cyber earlier this month.
The rest of the case rests on the defendants' own words. The complaint quotes a 2015 remark by Sam Altman that AI "will probably most likely lead to the end of the world," and Dario Amodei's estimate last year of roughly a 25% chance that AI goes "really, really badly." It also points to Anthropic's amended IPO filing, which the plaintiffs say warns that the harm from a catastrophic AI incident could exceed what the insurance industry can underwrite.
How the companies and courts have responded so far
A Meta spokesperson declined to comment, and the other defendants did not respond to requests for comment, according to CBS News and Reuters. No hearing date or ruling has been reported.
The suit adds to a growing set of court efforts to slow frontier work. Reuters reported that Florida, which sued OpenAI in June, asked a judge in September to bar the company from developing new models without outside oversight; that motion is still pending. After Florida's filing, an OpenAI spokesperson said the company had paused training its most capable models until it had additional safeguards. Eight states also sued OpenAI last month, seeking among other things an order against self-improving models that are not shown to be aligned.
Regulators are moving on a separate track. New York's RAISE Act registry requires frontier developers including OpenAI and Anthropic to register by Jan. 1, and CBS reported that Altman, Amodei and Elon Musk voiced support in October for an industrywide slowdown. The difference here is the remedy: a private UCL suit asking one state judge to set the limit for five companies at once. The defendants' first responses to the complaint will show whether they fight it on standing, on the scope of the Unfair Competition Law, or on the merits.



