OpenAI has been sued over the Hugging Face hacking incident, Axios reported on Sept. 29.

LASST (Legal Advocates for Safe Science and Technology), a public-interest legal group, filed a complaint against OpenAI in California Superior Court, the report said.

LASST, together with law firm Gerstein Harrow, asked the court to restrict OpenAI’s activities to prevent a recurrence of the hack.

The Hugging Face hack raised concerns that an AI agent could leave a testing environment and bypass safety guardrails.

There were also reports of tens of thousands of cases of agent actions that could be problematic. Against that backdrop, the lawsuit asks who bears legal responsibility when an AI agent goes beyond safety guardrails and causes real damage, Axios reported.

In its complaint, LASST argued that "OpenAI is responsible for the actions of its agents." It said OpenAI agents intentionally accessed Hugging Face without permission and that employees or executives knew of it or deliberately ignored it.

Tyler Whitmer (타일러 휘트머), the founder of LASST, stressed via Axios that the purpose of the lawsuit is to stop development practices in which AI agents independently cause harm outside.

He also said a legal mechanism is needed to determine the person or company responsible when an AI agent causes harm. He added he hopes an injunction will prompt OpenAI and the industry to change how they develop.

LASST is basing the lawsuit on California’s Unfair Competition Law. It argued that OpenAI conducted business illegally and unfairly, forcing time and costs to clean up the Hugging Face incident.

LASST also took issue with OpenAI turning off agent security measures and assigning tasks that could not be accomplished by prescribed methods. It said those practices violated California’s data access law and fraud-related laws.

Last year, California Governor Gavin Newsom signed a law designed to prevent defendants from avoiding responsibility by claiming that "AI acted on its own."