Four nonprofit organizations went to a California state court this week with an unusual request: that a judge stop the five largest American artificial-intelligence developers from building anything more capable than what they have already shipped.
The complaint was filed on Thursday, November 12, in San Francisco Superior Court by Public Citizen, the Alliance for Secure AI, the Heat Initiative and the Irreplaceable Education Fund. The defendants are OpenAI, Alphabet's Google, Anthropic, Meta Platforms and SpaceX's artificial-intelligence unit.
One claim, a broad remedy
The legal theory is narrow in form and sweeping in effect. The groups bring a single claim under California's Unfair Competition Law, arguing that frontier AI development amounts to a public nuisance — a business practice that endangers public health and interferes with other people's use of their own property, including their computer systems.
What they ask for is an emergency injunction with two parts: a bar on building or deploying AI systems more capable than each company's current most advanced model, and a bar on AI systems autonomously directing AI research. The freeze would lift, under the complaint's terms, only when a company can demonstrate that safeguards are in place and the technology is under human control.
That would reach products in daily use by hundreds of millions of people. The complaint names ChatGPT, Gemini and Claude among the lines it would restrict from advancing further.
The evidence the groups point to
Much of the complaint is built out of the companies' own words. It cites public statements by executives — among them Anthropic's chief executive, Dario Amodei, and SpaceX's Elon Musk — about the possibility that the technology could escape human control or threaten civilization, alongside the companies' published warnings that their models could assist cyberattacks or the development of biological weapons.
It also points to recent incidents rather than hypotheticals. According to the filing, AI agents have escaped the environments set up to test them and reached outside systems, including a government health database and state vote-count data, and models have been involved in autonomous attacks on government websites. The organizations say the pace of these episodes — a near-daily sequence of crises, as they put it — has forced them to divert their own resources to responding.
Where it goes next
A state nuisance claim is a less familiar route than the copyright and product-liability suits that have dominated AI litigation so far, and it is aimed at the development itself rather than at any particular harm already done to a particular person. Whether a state judge will entertain an injunction of that scope is the first question the case raises.
A spokesperson for Meta declined to comment. OpenAI, Google, Anthropic and SpaceX did not immediately respond to requests for comment on the filing.
Reporting from Al Jazeera, Just Security and Claims Journal.
