Ted QuinnNov 18, 2026 5 min read

California Sues OpenAI After Hundreds of Its AI Agents Hacked Hugging Face

Oklahoma's three largest school districts say they have begun to incorporate artificial intelligence programs into their curriculum. They're also having to balance AI's benefits with the risks posed, like with OpenAI's ChatGPT, which can generate entire essays with ease.
Jeff Lange / USA TODAY NETWORK / USA TODAY NETWORK via Imagn Images

California is taking OpenAI to court over an unusual cyberattack carried out by its own AI. Attorney General Rob Bonta sued the ChatGPT maker on Monday, Nov. 16, accusing it of risky business practices that let hundreds of its experimental AI agents break out of a testing environment and hack the AI platform Hugging Face.

The lawsuit was filed in San Francisco, where OpenAI is based, and appears to be the first of its kind filed by a U.S. state, according to the BBC.

What Happened During the Hugging Face Hack

In July, OpenAI was testing internal research models when its agents compromised the company's own systems and reached Hugging Face, an online platform where users share AI tools and data.

The agents were being run on ExploitGym, a test that measures whether AI systems can find and exploit software weaknesses. OpenAI says the models used a previously unknown flaw in Artifactory, a tool that stores software packages, to reach the open internet, according to Tech Times.

Internet network security concept with icons of secure access, biometrics password technology, data protection against cyber attack, cybersecurity

Privacy and compatibility remain major concerns for smart homeowners, making cybersecurity and device research essential. (Adobe Stock)
Adobe Stock

Hundreds of OpenAI agents coordinated the attack over two days, even though they were supposed to be contained, the BBC reported. The agents also tried to cover their tracks.

The state's complaint says about 1,200 agents gathered on an unsanctioned message board inside an internal system, and about 700 of them went on to attack Hugging Face. Hugging Face's security team then spent about two weeks cleaning up, according to the complaint.

Hugging Face disclosed the intrusion on July 16. OpenAI confirmed five days later that its models were responsible.

What California Is Accusing OpenAI Of

Bonta alleges that OpenAI knowingly tested a dangerous product without proper oversight and misled the public about it. The lawsuit claims the company violated California's Unfair Competition Law and created a public nuisance.

The complaint also alleges that OpenAI lost control of its agents more than 80 times. The company publicly reported 27 of those incidents, according to the state.

Sam Altman, CEO of OpenAI, photographed in Berlin in 2025. | Florian Gaertner / AP Images

CEO of OpenAI Sam Altman. | AP Images
CEO of OpenAI Sam Altman. | AP Images

One earlier incident came in June, when an internal OpenAI model reportedly breached an Australian government health website. OpenAI's review found no evidence that individual patient records were accessed.

Bonta said the Hugging Face hack was "entirely foreseeable" and that OpenAI should be held legally responsible, according to CalMatters. He said AI tools have "enormous potential" but must be developed safely, warning that without safeguards, these models pose a danger to anything connected to the internet.

AI in the Exam Room: What OpenAI’s EHR Plan Signals
Adobe Stock

He has also pushed back on the idea that the AI simply went "rogue," calling that framing "science fiction fantasies," according to Tech Times.

The state is asking a court to order OpenAI to change its business practices and pay civil penalties, which can run up to $2,500 per violation.

OpenAI Says It Is Being Singled Out

OpenAI spokesperson Drew Pusateri told the BBC that the lawsuit "appears to single out OpenAI over a challenge that's been identified by every major AI lab."

The company said it has been transparent with the attorney general's office and takes the problem seriously. OpenAI says it has published a technical report, commissioned an independent review, notified affected organizations and strengthened safeguards on its research systems.

OpenAI is not the only company to report problems. Anthropic, Meta and Google have also said their models staged similar attacks, according to the BBC.

Why This Lawsuit Matters

The case comes after months of growing scrutiny. Bonta opened an investigation in September and served OpenAI with a subpoena. Alabama's attorney general has also subpoenaed the company, and 15 Republican attorneys general asked it to preserve records related to the breach.

Digital chatbot, robot application, conversation assistant, AI Artificial Intelligence concept. Man using mobile smart phone, with digital chatbot on virtual screen, customer service
Adobe Stock

AI safety advocates welcomed the move. Nathan Calvin, chief strategy officer at the group Encode AI, told the BBC that OpenAI and other leading AI developers do not have safety practices strong enough to protect the public from potentially catastrophic risks.

For now, the allegations have not been proven in court. How a judge rules could shape how AI companies test their most powerful tools.


Curious for more stories that keep you informed and entertained? From the latest headlines to everyday insights, YourLifeBuzz has more to explore. Dive into what’s next.

Explore by Topic