The lawsuit’s request is straightforward: for the court to prohibit the defendants from continuing to develop or deploy new AI models.
The plaintiffs, four U.S. nonprofits, sued Thursday in San Francisco Superior Court. The defendants are named OpenAI, Anthropic, Google, Meta, and SpaceX’s AI business.
The legal weapon they used was California’s Anti-Unfair Competition Law. The reason is that these companies do not have the ability to effectively prevent "catastrophic" damage to their own systems, and continuing to promote the development of cutting-edge models will endanger public health and interfere with the public's normal use of property such as computer systems. The basis cited in the complaint is cases of AI involvement in cyber attacks that have occurred in recent years.
From a legal perspective, this path is difficult to navigate. This California law usually governs unfair or deceptive business practices. To use it to request the suspension of research and development of an entire industry, the court needs to first determine that "training new models" itself is an unfair business practice, and the threshold is very high.
But the role of such litigation is often not in adjudication. The filing of the case itself will require the defendant to hand over materials such as internal safety assessments and accident records during the discovery stage, and public pressure will follow. The timing is also coincidental: Anthropic will be listed next week, and there may be one more paragraph of risk warning in the prospectus.
Most of the lawsuits surrounding AI in the past two years have been copyright disputes. This time, they are directly focused on "whether we can continue to do it." This is a change in the direction of litigation.