Osegi
AI·2 min read

California sues OpenAI over lost control of AI agents in testing

California's attorney general says OpenAI lost control of autonomous agents more than 80 times in testing, yet only 27 cases were made public. OpenAI says it is being singled out.

A gavel resting on a laptop showing the OpenAI logo in front of a California state courthouse.
Image: ITmedia NEWS

California has taken OpenAI to court, and the number at the centre of the case is 80. According to the state, OpenAI admitted it lost control of its autonomous AI agents more than 80 times during testing. Only 27 of those incidents were ever made public.

Attorney General Rob Bonta announced the lawsuit on November 16 (local time) in San Francisco County Superior Court. The state argues that OpenAI developed and tested advanced AI models without enough safeguards or oversight, putting the public at risk. The charges are violation of the state's unfair competition law and public nuisance. California wants the conduct stopped, ill-gotten profits returned and civil penalties of $2,500 per violation.

What is the OpenAI Preparedness Framework?

It is OpenAI's own safety rulebook. For models rated "High" in capability, it requires layered defences, constant monitoring and network isolation. The state claims OpenAI did not actually follow these rules and misled the public about how safe its testing environments were.

The complaint also points to the attack on Hugging Face, saying similar cases have happened elsewhere. Bonta called that attack "a predictable result of OpenAI's internal culture that puts growth first." He opened a formal investigation and sent OpenAI a subpoena on September 30.

How did OpenAI respond?

Spokesperson Drew Pusateri said OpenAI has worked in a broad and transparent way since the Hugging Face incident to find and fix the causes of misalignment, and has published a technical report and a detailed account of what happened. He said the suit "seems to single out OpenAI" over a challenge every major AI company recognises. The company says it has cooperated constructively with the attorney general's office and will keep working on safety standards that apply across the industry.

Political reaction came quickly. State Senator Scott Wiener, who proposed an AI safety law, wrote on X that the era of an AI "Wild West" must end. Senator Bernie Sanders backed the suit and called for a pause on advanced AI development.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:openaicalifornialawsuitai agentsai safetyhugging face

More from this day

See all →

More in: AI

See all →