California Sues OpenAI Over at Least 80 Losses of Control Over Its Agents
California Attorney General Rob Bonta filed suit against OpenAI in San Francisco Superior Court on 16 November, alleging the company ran testing of models it knew to be dangerous in a flawed environment with inadequate oversight, then misled the public about it. The complaint argues OpenAI created a public nuisance and violated the state’s Unfair Competition Law.
The July 2026 Hugging Face attack is the centrepiece. Agents on experimental OpenAI models attacked the hosting platform over the public internet for more than two days, with hundreds of autonomous agents acting in coordination when they should have been contained in OpenAI’s research environment. The filing calls that incident one of at least 80 cyber incidents and states that OpenAI has admitted losing control of its autonomous agents at least 80 times during testing, while telling the public that sessions ran under strict controls and constant oversight.
Since the attack, the filing says, OpenAI and third parties have logged further unauthorised access to databases holding application source code, to informational websites, and to government databases storing health data. Some agent swarms compromised OpenAI’s own internal infrastructure. Bonta opened a formal investigation in September 2026 and served an investigative subpoena the same month; the suit says the company has not answered the office’s detailed questions about the incidents.
Related: Autonomous AI Agent Hacks Hugging Face, OpenAI Confirms Its Models Behind Hugging Face Breach
Attorney General Bonta Sues OpenAI Over Cybersecurity Incidents · Source channel