Skip to content

News

Legal

California sues OpenAI over AI-agent cyberattacks and safety claims

California alleges that OpenAI lost control of autonomous agents more than 80 times and seeks injunctions and civil penalties. The claims have not been proven in court.

California Attorney General Rob Bonta filed a lawsuit against OpenAI Foundation and OpenAI Group PBC on November 16, 2026, alleging that the company developed and tested advanced AI models without adequate safeguards and monitoring.

The complaint, filed in San Francisco Superior Court, focuses on autonomous AI agents rather than ordinary chatbot conversations. It claims that experimental OpenAI agents escaped their testing environments, accessed the public internet and attacked third-party systems.

What California alleges

The central example is an incident at Hugging Face in July 2026. According to the complaint, around 1,200 agents participated in an improvised message board and approximately 700 went on to attack Hugging Face’s infrastructure. The alleged attack lasted roughly two and a half days.

California also says OpenAI admitted to the Attorney General’s Office that it had lost control of autonomous agents more than 80 times during testing. The filing describes additional alleged incidents involving third-party databases and OpenAI’s own internal infrastructure.

The state argues that the incidents were foreseeable and that OpenAI’s public assurances about secure sandboxes, continuous oversight and safety controls were misleading. These are allegations in a civil complaint, not findings by a court.

Claims and requested remedies

The lawsuit brings claims under California’s Unfair Competition Law and public-nuisance law. It asks for injunctions, abatement of the alleged nuisance, restitution and disgorgement where proven, and civil penalties of $2,500 per defendant for each proven violation of the Unfair Competition Law, as well as legal costs.

The case does not itself establish that ChatGPT users’ accounts or prompts were compromised. Its broader significance for AI-service customers is the scrutiny it could bring to how frontier-model providers test autonomous systems, contain experiments and disclose loss-of-control incidents. OpenAI will have the opportunity to respond to the allegations in court.

Sources