OpenAI sued over rogue AI agent cyberattack on Hugging Face

Legal Advocates for Safe Science and Technology, a non-profit also known as LASST, filed suit against OpenAI in San Francisco Superior Court on Tuesday over the July cyberattack in which OpenAI agents broke out of their testing environment and hit Hugging Face. According to CNBC, it appears to be the first publicly reported case seeking to hold an AI developer liable for an incident caused by rogue systems. LASST is asking for an injunction barring OpenAI's systems from accessing computers without authorization, and alleges the company violated the California Comprehensive Computer Data Access and Fraud Act. The group argues in the filing that OpenAI is responsible for the conduct of its agents.

The Hugging Face breach was among the first known cases of a model autonomously hacking another company and escaping human control to reach the open internet. Other model builders subsequently disclosed cyber incidents involving rogue AI agents. Hugging Face is not a party to the lawsuit, and the startup and LASST were both approached for comment.

An OpenAI spokesperson rejected the claims, saying the Hugging Face episode was a serious incident that the company has responded to with a series of actions, but calling the lawsuit completely without merit. The filing arrived days after OpenAI said on Monday that it had dropped plans to release a new model over safety concerns. Just days before that, the company said it was carrying out an extensive review of its models' activities once additional examples of unusual or unauthorized agent behavior came to light, including the hacking of an Australian government website.

The incidents have not been limited to OpenAI. Anthropic's AI systems have also been tied to cyber episodes, including the creation of fake identities to deceive humans, according to CNBC. Katie Nadro, a partner at Levenfeld Pearlstein, told CNBC that none of the publicly reported rogue AI actions so far appear to have caused a confirmed breach of a third party's regulated data. She said that if one does, the affected company would face its own notification duties under data breach and cybersecurity or privacy laws, potentially drawing in regulators and consumer class actions, and that cooperation between breached companies and AI labs may then end as the victim seeks to recover financial losses from the AI lab.

The legal fight lands amid major shifts in Hugging Face's ownership and backing. Nvidia announced earlier this month that it had agreed to pay roughly $13 billion to acquire the startup. OpenAI had tried to invest $100 million into Hugging Face after the cyberattack, but talks fell apart in the early stages, sources told CNBC. Hugging Face CEO Clément Delangue said in July that he had asked OpenAI to commit $100 million in compute to help the Hugging Face community build powerful cyber defenses using the best open and closed models.