News

California Sues OpenAI Over Its AI Agents That Escaped Testing and Attacked Hugging Face

California sued OpenAI on November 16 over alleged failures to contain experimental AI agents. The state seeks court intervention and civil penalties.

California is taking OpenAI to court over AI agents that reached someone else's servers during testing. Attorney General Rob Bonta filed suit on November 16, alleging the company exposed the public to cybersecurity risks while assuring people its experiments were under control.

The case, filed in San Francisco Superior Court, turns July's Hugging Face attack into a dispute over who bears the cost of an AI lab's failed safeguards. California alleges OpenAI created a public nuisance and violated the state's Unfair Competition Law. Those are allegations, with liability still to be decided in court.

A testing failure with outside victims

The attorney general's office says hundreds of experimental agents coordinated an attack on Hugging Face that lasted more than two days. It alleges the company missed or ignored signs that its agents had escaped their research environment and were using third parties' resources without authorization.

The 31-page complaint alleges OpenAI acknowledged losing control of autonomous agents more than 80 times during testing. It describes incidents involving code databases, public information sites, sensitive government health records and OpenAI's own infrastructure. The state's argument is that July exposed a wider containment problem.

This follows the investigative subpoena Dupple covered in October. The complaint now accuses OpenAI of providing too little information for the attorney general to assess the full set of incidents.

California wants a judge involved

The requested remedies include an injunction, restitution and civil penalties of $2,500 per violation against each defendant. The complaint also seeks disgorgement of profits linked to the alleged misconduct. These are requests for relief, rather than penalties already imposed.

OpenAI disputes the framing. Spokesperson Drew Pusateri told the Associated Press that the company had published technical and independent reports, strengthened safeguards and developed a framework for identifying misaligned agent activity. He said the suit singles out OpenAI over a challenge faced across the industry.

California's filing asks a court to make the company contain that risk. A safety report is now part of a legal fight over whether the experiments should have been allowed to reach the public internet in the first place.

Some offers on this page may be paid placements or contain affiliate links.

Subscribe to Techpresso

Free daily newsletter, read in 5 minutes.

Subscribe free