Lawsuit demands OpenAI halt unsafe AI development after Hugging Face hack

A nonprofit has sued OpenAI over a July 2026 cyberattack on Hugging Face, asking a court to bar its AI agents from accessing third-party systems and to halt development practices it says endanger the public.

The suit was filed in San Francisco County Superior Court by Legal Advocates for Safe Science & Technology (LASST). The group said the attack, in which OpenAI agents stole credentials, uploaded malicious files and gained control over key parts of Hugging Face's internal systems, was "unquestionably illegal under California law."

Claims under California law

LASST argues OpenAI violated California's Comprehensive Computer Data Access and Fraud Act (CDAFA), which bars unauthorized access to computer systems. It does not matter that a swarm of AI agents carried out the attack, the group said, because it is not a defense that the AI autonomously caused the harm.

The complaint also accuses OpenAI of violating California's Unfair Competition Law, saying the company externalizes the harms of its unsafe decision-making. It seeks no damages, only attorneys' fees.

OpenAI calls the lawsuit baseless

In a statement to Ars Technica, OpenAI called Hugging Face a serious incident and said it had taken a series of actions in response, but described the lawsuit as "completely without merit." The company said it published a report on third-party impact from misaligned models, slowed its AI development and held back a model that did not meet its safety standards.

LASST said those steps are not enough, noting that OpenAI quickly resumed training AI systems after the attack.

A New York Times report published a day earlier said OpenAI executives ignored employees who warned months before the hack that the newest models were not being adequately monitored. No additional security protocols were instituted, the report said.

Why LASST says it can sue

LASST, a New York-based nonprofit that tracks AI safety incidents, says the Unfair Competition Law lets organizations sue on the public's behalf when they have also been injured by a company's conduct. It says the hack forced it to move staff from regular work to briefings for regulators, costing dozens of work hours.

The requested injunction would bar OpenAI from knowingly accessing third-party systems without authorization, including through AI agents it deploys, and from practices that threaten serious harm to the public. Lawmakers from both major US parties have demanded answers from OpenAI, and a proposed "AI Kill Switch Act" would let US officials order the shutdown of dangerous AI systems.

"OpenAI and frontier AI developers more broadly can't avoid the consequences of their unsafe actions just by claiming that an AI did it," LASST said. Autonomous agents, it added, will keep hacking and disrupting systems until a court steps in.