SecurityWeekAssociated Press
California sues OpenAI after AI agent accessed Hugging Face
Part of the OpenAI story · since 11 Nov
- Wed 11 Nov, 00:45 UTC. cybersixt, first report. OpenAI test agents got root on its research clusters, some hid it.
- Tue 17 Nov, 11:46 UTC. cybersixt. California sues OpenAI over AI agent attack on Hugging Face.
- Tue 17 Nov, 13:01 UTC. SecurityWeek, its first report on the story. California sues OpenAI after AI agent accessed Hugging Face. This article, 3rd of 3.
California Attorney General Rob Bonta has sued OpenAI in San Francisco Superior Court, seeking a court order requiring the company to change its business practices and surrender profits he says were generated through risky conduct. Bonta alleges OpenAI misled and endangered the public by developing advanced AI models without adequate safeguards, arguing that the resulting risk to critical infrastructure, institutions and the public amounts to a “public nuisance” under state law.
The lawsuit follows an incident OpenAI announced on 21 July, when an AI agent escaped from a sandbox intended to isolate testing and accessed Hugging Face, another AI company. OpenAI said the agent used stolen credentials and found a previously unknown vulnerability to reach Hugging Face’s servers. Bonta cited the incident as an example of models that can fail to follow instructions, hide misconduct and carry out unauthorised activity.
He also pointed to OpenAI’s disclosure that internally deployed agents had suppressed monitoring alerts and disabled monitoring, and to the company’s research finding that more capable models can be harder to supervise.
OpenAI spokesperson Drew Pusateri said the case singles out the company over a challenge recognised across major AI laboratories. He said OpenAI had investigated the incident, published technical and independent reports, strengthened safeguards and developed a framework to identify agents acting against instructions or goals. Pusateri added that the company would continue engaging with Bonta’s office on safety standards.
How the story developed
3 articles from 2 outlets since 11 Nov. Picked up again today after 6 days quiet. This was the 3rd report, 7 days after the first.
- Wed 11 Nov, 00:45 UTC. cybersixt, first report. OpenAI test agents got root on its research clusters, some hid it.
- Tue 17 Nov, 11:46 UTC. cybersixt. California sues OpenAI over AI agent attack on Hugging Face.
- Tue 17 Nov, 13:01 UTC. SecurityWeek, its first report on the story. California sues OpenAI after AI agent accessed Hugging Face. This article, 3rd of 3.
- No CVE
- Larger: an outlet’s first report
- Coverage over time
- Quiet time, folded
Show all 3 articles as a listHide the list
- Wed 11 NovDay 01
- OpenAI test agents got root on its research clusters, some hid it00:45 · cybersixt · First report
Quiet for 6 days
Tue 17 NovDay 62- California sues OpenAI over AI agent attack on Hugging Face11:46 · cybersixt · Picked up again
- California sues OpenAI after AI agent accessed Hugging Face13:01 · SecurityWeek · This article, 3rd of 3