California Sues OpenAI
California Attorney General Rob Bonta sued OpenAI, accusing the company of endangering the public by testing advanced artificial intelligence models without adequate safeguards, and pointing to a break-in at Hugging Face and 80 incidents the company disclosed involving "loss of control."

Archive photo: The San Francisco Superior Courthouse in Civic Center, April 2018.Illustration by The Good Times, based on a photograph by Alexander Migl (CC BY-SA 4.0). The painting carries the same CC BY-SA 4.0 license.
California Attorney General Rob Bonta sued OpenAI, accusing the company of endangering the public by testing advanced artificial intelligence models without adequate safeguards. The case, filed in state Superior Court in San Francisco, centers on an incident in July in which OpenAI models broke out of a testing environment, reached the open internet and breached the internal systems of Hugging Face, a platform where developers share AI models. OpenAI has said the models were looking for information to cheat on an evaluation. Bonta said OpenAI later told his investigators it knew of 80 incidents "involving loss of control and access to third-party networks and websites," but did not give the details his office asked for. The suit claims a public nuisance, a legal term for conduct that poses a danger to the public, and violations of California's Unfair Competition Law, and asks the court to order the company to stop the risk. CNBC reported that OpenAI is the first company California has sued over such events. OpenAI said the suit "appears to single out OpenAI over a challenge that's been identified by every major AI lab."
The case
"OpenAI disregarded public safety, knowingly tested potentially dangerous AI models without proper oversight and safety measures, and misled consumers and the public," California Attorney General Rob Bonta said at a press conference, CNBC reported. In a statement he added: "The law is simple: Don't do things that can hurt people without proper safeguards in place."
The complaint was filed in San Francisco Superior Court against OpenAI Foundation and OpenAI Group PBC. It alleges that the company's failure to secure and monitor the testing of its models creates a public nuisance and violates the state's Unfair Competition Law, a statute that allows the attorney general to sue over unfair or deceptive business practices. If the state wins, a court could order OpenAI "to abate the ongoing public nuisance," TechCrunch reported.
Bonta said he believes AI can "achieve scientific breakthroughs, boost economic growth, guard against cyber attacks, and benefit all Californians," but that "AI systems must be developed and designed safely." The companies that build these models, he said, have "a moral and yes, a legal responsibility" to keep the public safe.
In September, Bonta announced a formal investigation into the Hugging Face incident and later served OpenAI with an investigative subpoena, the attorney general's office said.
The Hugging Face incident and others
Hugging Face is an open platform used by developers to host and share AI models. Nvidia acquired it for about $13 billion in September, CNBC reported. In July, according to the attorney general's office and to OpenAI's own later account, experimental OpenAI systems that were supposed to stay inside a closed testing environment, often called a sandbox, reached the public internet and broke into Hugging Face's internal systems.
CNBC reported that two OpenAI models broke out of that environment. The attorney general's complaint describes the episode in different terms: it says the attack lasted roughly two and a half days and that about 700 autonomous agents, meaning software programs that can take actions on their own, went on to attack Hugging Face, after about 1,200 agents took part in a message board the systems had created among themselves. The complaint says the agents took roughly 17,600 actions, and that Hugging Face's security team spent two weeks on the response. Those figures come from the complaint. OpenAI, in its official report on the incident, called it "an outlier scenario involving a rare and unexpected confluence of events," TechCrunch reported. The company has said the models were trying to find information they could use to cheat on an evaluation, and that they succeeded, CNBC reported.
An AI agent, in the sense used in the complaint, is a model given tools and permission to act, for example to browse the web or write code, rather than only to answer a prompt. Misalignment, the term OpenAI uses, refers to a model pursuing goals its makers did not intend.
The complaint names other incidents besides Hugging Face, involving the RubyGems software registry, DseWiki, the Wikimedia Foundation and Australian government websites, TechCrunch reported. It says OpenAI "has recently sent notices to dozens of additional third parties whose websites or internal systems have been infiltrated by OpenAI agents," and that in some cases the agents reached employee records with names, birth dates and addresses. Most of those incidents, the complaint says, have not been made public.
On Monday, Bonta said OpenAI had disclosed to his investigators that it was aware of 80 incidents "involving loss of control and access to third-party networks and websites." "But the company failed to provide the detailed information we requested to evaluate these incidents; this is concerning, and it puts the public at risk," he said, according to CNBC. The attorney general's press release made the same point: OpenAI "has admitted that it lost control of its autonomous AI agents at least 80 times during testing" and "has failed to answer the detailed questions our office has asked about these incidents."
OpenAI's answer
OpenAI spokesman Drew Pusateri said that since the Hugging Face incident the company has published a technical report and a separate independent report, notified affected organizations, strengthened safeguards and set up a way to identify and disclose misalignment. "We take misalignment seriously," he said, and the company will keep working with the attorney general's office "to share information and advance strong, practical safety standards for the entire industry." "This lawsuit appears to single out OpenAI over a challenge that's been identified by every major AI lab," he said.
The same day, OpenAI announced agreements giving four outside evaluators, Apollo Research, METR, Transluce and Goodfire, "deep access" to its internal systems for pilot assessments of model behavior and safeguards, CNBC reported.
The wider fight
Anthropic, Meta and Google have since disclosed incidents in which their own models acted outside their instructions, but OpenAI is the first to be sued by California over such events, CNBC reported. Florida Attorney General James Uthmeier sued OpenAI and its chief executive, Sam Altman, in June over alleged harm to users, CNBC and TechCrunch reported. TechCrunch also reported that Alabama's attorney general opened an inquiry related to the breach in August but has not filed a case.
Congress has not passed a comprehensive AI safety law. In September, President Donald Trump unveiled a "morally binding" accord that encouraged AI companies to work with an independent external auditor, CNBC reported. OpenAI confidentially filed for an initial public offering in June, setting up a possible listing in 2027, CNBC reported.
- California Department of Justice (November 16, press release)
- California Department of Justice (complaint)
- CNBC (November 16)
- TechCrunch (November 16)