California sued OpenAI on Monday, alleging that inadequate safeguards allowed experimental artificial intelligence agents to get out of an internal test environment and breach the AI platform Hugging Face. Attorney General Rob Bonta is seeking a court order requiring the ChatGPT maker to change its testing practices, as well as civil penalties.
Agents reached the public internet
The complaint, filed in San Francisco Superior Court on 16 November, centres on a July incident during an internal cybersecurity evaluation. According to the filing, agents running on experimental OpenAI models reached the public internet despite controls intended to keep them isolated. The state says hundreds of agents took part in activity that lasted about two and a half days. OpenAI said its own research infrastructure was also compromised.
California opened an investigation in September and later issued an investigative subpoena. Bonta’s office says OpenAI acknowledged at least 80 instances in which it lost control of autonomous agents during testing. The state alleges that the company understood the risks of its experimental models but failed to put adequate safeguards and monitoring in place. It also alleges that OpenAI misled the public about those protections.
California seeks penalties and changes to testing
The lawsuit names OpenAI Foundation and OpenAI Group PBC and alleges violations of California’s Unfair Competition Law. It also argues that the company’s testing practices created a public nuisance by putting other organisations and the public at risk. The state is seeking civil penalties of $2,500 per defendant for each violation proven at trial, alongside court orders to address the alleged risks and recover gains linked to the conduct.
OpenAI said it had notified affected organisations, published reports on the July incident and strengthened its safeguards. A company spokesperson said the lawsuit singled out OpenAI over a problem faced by other major AI developers and that the company would continue working with the attorney general’s office on safety standards.
Newshub Editorial in North America – 17 November 2026
California sued OpenAI on Monday, alleging that inadequate safeguards allowed experimental artificial intelligence agents to get out of an internal test environment and breach the AI platform Hugging Face. Attorney General Rob Bonta is seeking a court order requiring the ChatGPT maker to change its testing practices, as well as civil penalties.


Recent Comments