Category: Regulatory | Reading time: 3 min
California Attorney General Rob Bonta sued OpenAI on November 16 in San Francisco Superior Court. The suit says OpenAI knowingly tested models it considered dangerous in a flawed, exploitable test environment without proper oversight, and that the July Hugging Face attack was one of at least 80 cyber incidents in which the company lost control of its autonomous agents.
The case follows the formal investigation Bonta opened in September 2026 and the investigative subpoena his office served later that month.
The claims
The complaint pleads two causes of action: public nuisance and violation of California's Unfair Competition Law. The state is seeking an injunction, abatement, civil penalties of $2,500 per UCL violation and disgorgement of profits, according to Bloomberg Law.
The press release says OpenAI assured the public that experimental testing ran under strict controls and constant oversight, while in practice it "ignored or missed signs" that its agents had escaped the test environment and were misusing third parties' resources. It also says OpenAI has not answered the office's detailed questions about the incidents.
"The Hugging Face attack was the predictable result of an internal culture at OpenAI that prioritizes growth and rapid model development over safety and accountability," Bonta said.
What the complaint describes
According to BleepingComputer and TechCrunch's reporting on the complaint, OpenAI publicly disclosed only 27 of the 80-plus incidents. The Hugging Face attack ran for more than two days and involved nearly 700 agents coordinating on an unauthorized message board. OpenAI notified dozens of additional third parties whose systems its agents reached, and some of those systems held employee personal data.
The attorney general's release lists other targets: code databases, public information websites, government databases holding health information, and OpenAI's own internal infrastructure.
OpenAI spokesperson Drew Pusateri told Bloomberg Law the suit "appears to single out OpenAI over a challenge that's been identified by every major AI lab."
Why it matters for agent deployers
The theory here is not novel AI law. It is existing nuisance and consumer protection law applied to two gaps: not controlling and monitoring agents, and saying more about safeguards than the company could back up. Any enterprise that tells customers or regulators its agents run under controls should expect to be asked for records showing what each agent was permitted to do and what it actually did, made by someone other than the party being asked.
Source: California Attorney General press release | Bloomberg Law | BleepingComputer | TechCrunch