Nonprofits ask a San Francisco court to freeze AI where it is, using a law written for unfair business

The biggest names in artificial intelligence have been warned for years that their technology could one day get out of hand. Now a group of nonprofits is asking a judge to make them stop where they are.

The coalition, which includes the Alliance for Secure AI, Public Citizen, Heat Initiative and the Irreplaceable Education Fund, filed its complaint on Thursday in San Francisco County Superior Court. The defendants are OpenAI, Alphabet’s Google, Anthropic, Meta Platforms and SpaceX’s AI division, whose products include ChatGPT, Gemini and Claude.

What exactly do they want?

An injunction barring the five from developing or deploying any system more capable than their most advanced current models. The groups also want a ban on AI systems built to carry out or direct the research and development of future models on their own, the step many researchers see as the point where humans could lose control.

Brendan Steinhauser, chief executive of the Alliance for Secure AI, said the coalition was asking the court to hold AI companies to the same standard as other industries, which are not allowed to keep building products that may be unsafe.

How can a business law stop AI?

California’s Unfair Competition Law allows people and organisations to sue to stop business practices they say are unlawful or harmful. The coalition argues that continuing to build more powerful AI is “substantially injurious” if it leads to disaster, and that it is a public nuisance because others must spend money hardening their systems against AI-driven hacking.

The suit also argues the companies’ conduct conflicts with SB 53, the California law aimed at reducing catastrophic risks from the most powerful AI models.

To make its case, the complaint leans on the companies’ own words. It cites comments by leaders including Anthropic chief executive Dario Amodei and SpaceX chief executive Elon Musk about AI’s potential to destroy civilisation. It also describes incidents in which, it says, AI models escaped testing environments and attacked websites on their own.

Will it work?

That is the open question. A central test is whether organisations that have not been directly harmed by an AI product can sue to stop it being built. The case is unusually broad: it targets five companies at once, and systems that do not yet exist.

The claims in the complaint are allegations and have not been tested in court.

What have the companies said?

A Meta spokesperson declined to comment. The other companies did not immediately respond to requests for comment.

Why does it matter here?

Almost every Australian who uses an AI chatbot, at work, at school or on their phone, is using a product from one of these five companies. An order freezing their most advanced work would ripple straight into the tools Australian businesses are building on.

It also lands as governments split on how hard to regulate. Washington has so far declined to pursue large-scale AI regulation, and Australia’s federal government decided in December 2025 to rely on existing laws rather than a standalone AI act. With legislators moving slowly, campaigners are increasingly turning to the courts.