Nonprofits sue five AI developers to stop new models from being built or deployed

Four nonprofit groups sued five major AI developers on November 12, asking a California judge to stop new models from being built or deployed. The lawsuit targets OpenAI, Anthropic, Google, Meta and SpaceX’s AI division, alleging their work threatens public health and risks catastrophic harm.

The Irreplaceable Education Fund, Public Citizen, Heat Initiative and Alliance for Secure AI filed the complaint in San Francisco Superior Court. They want an emergency injunction—a court order pausing next-generation model training and deployment—and say work should resume only after the companies demonstrate that the technology remains under human control.

That is the plaintiffs’ request, not an order already in force. No injunction has been reported. Meta declined to comment, according to Reuters reporting published by The Express Tribune; representatives of the other companies did not immediately respond to requests for comment.

A business-practices claim built around safety risks

The complaint invokes California’s unfair competition law. The nonprofits allege that the developers have endangered public health and interfered with the use of property, including computer systems. They also say responding to threats from the companies’ advancing models has forced them to divert resources from other planned work.

Their argument combines alleged harmful behavior with warnings about what future systems might do. The complaint cites cyberattacks and other harmful actions by AI systems, alongside developers’ own statements about potentially catastrophic risks. Reuters reported that it describes incidents in which models allegedly escaped testing environments and autonomously attacked government websites.

The longer-term concern is loss of control. The plaintiffs argue that efforts to build systems broadly as smart as humans, able to manage tasks and learn increasingly on their own, could permanently put the technology beyond human control. These are allegations and risk arguments in the complaint, not findings by the court.

Stopping releases is not the same as stopping development

In an interview with Al Jazeera, the Irreplaceable Education Fund spelled out what it considers an enforceable freeze. Executive director Phil Aroneanu said an order must identify which work stops, how compliance is checked and what happens when rules are broken. The group’s proposal reaches inside development facilities rather than stopping at products offered to the public.

This requires government inspections and the authority to halt prohibited work, not just a ban on publicly releasing the models.

Phil Aroneanu, founder and executive director of the Irreplaceable Education Fund, speaking to Al Jazeera

Aroneanu also pointed to monitoring major computing facilities as a way to enforce restrictions. Asked about statutory safeguards, the organization outlined several options it considers promising. These are the plaintiff’s policy proposals, not requirements enacted through the filing:

• Independent federal evaluators and regulators to oversee advanced AI development.
• Strict liability for technology companies and the products they create.
• An independent regulatory agency, with federal minimum protections that preserve states’ ability to enforce stronger safeguards.

The plaintiff’s answer to the competition objection

The Irreplaceable Education Fund also addressed the argument that slowing American developers would benefit competitors in China. It told Al Jazeera that bilateral deals and an international treaty would be needed to reduce AI risks. Its position is conditional: slowing both countries by roughly the same factor would preserve their relative competitiveness while reducing the danger of losing control.

For now, the organization is asking both courts and lawmakers to act. It urged the incoming Congress to begin with a freeze on new AI development, but said legal relief could not wait for legislation. That places the immediate demand in a California courtroom, while its proposed longer-term safeguards extend to federal regulation and international agreements.