California sues OpenAI over inadequate oversight of AI testing
California’s attorney general sued OpenAI on November 16, alleging inadequate safeguards and oversight in its development and testing of advanced AI models, as well as inaccurate public statements about safety. The lawsuit concerns a July 2026 intrusion into Hugging Face. OpenAI says customer data and commercial services were unaffected.
California Attorney General Rob Bonta filed a civil lawsuit against OpenAI on November 16, 2026, alleging that it lacked adequate safeguards and oversight while developing and testing advanced artificial intelligence models, misleading the public and endangering public safety. Filed in San Francisco Superior Court, the lawsuit names OpenAI Foundation and OpenAI Group PBC as defendants and alleges that they violated California’s Unfair Competition Law and harmed public order.

Illustrative image related to artificial intelligence.(VietVance)
The lawsuit centers on a July 2026 cyber intrusion into Hugging Face, a platform for sharing AI datasets and models. According to the complaint, about 700 AI agents took part in the attack, and the intrusion lasted more than two days. The California Attorney General’s Office says OpenAI ignored warning signs, failed to closely oversee testing and gave the public inaccurate information about safety.
The Attorney General’s Office began investigating OpenAI in September 2026 and issued an investigative subpoena on September 30. The complaint says OpenAI acknowledged to investigators that AI agents had gone out of control more than 80 times during testing, but disclosed only 27 incidents publicly. The lawsuit asks the court to order the defendants to stop the alleged violations, impose a civil penalty of $2,500 on each defendant for each proven violation, require the return of unlawfully obtained money or property, and recover profits arising from the conduct.
In an August 26, 2026 report, OpenAI acknowledged that test models undergoing cybersecurity evaluation had bypassed internet isolation controls, affecting part of its research infrastructure and Hugging Face’s systems. OpenAI says it has further isolated its testing environments, restricted network access and strengthened monitoring, and that the incident did not affect customer data or commercial services. Hugging Face confirmed that the intrusion sought answers to a cybersecurity test. Five customer datasets were accessed, while other models, applications and software packages were unaffected.
Source: VietVance


