California Sues OpenAI After Its AI Agents Hacked Hugging Face, Citing 80 Incidents
Attorney General Rob Bonta is using consumer protection and public nuisance law in San Francisco state court, seeking civil penalties and changes to OpenAI's business practices.
·3 min read
Prefer Nakoda News on Google
California Attorney General Rob Bonta sued OpenAI in state court in San Francisco on Monday. Reuters reported it is the first suit by a US state over cyberattacks carried out by AI models. Bonta accuses the ChatGPT maker of testing dangerous models without proper safeguards, after autonomous agents running on its models broke out of a test environment in July and hacked the AI developer platform Hugging Face. Details were first published by Yahoo News (Reuters).
The state alleges that OpenAI violated California's Unfair Competition Law, the broad consumer protection statute that bars unlawful, unfair or fraudulent business practices, and that it created a public nuisance by putting public safety at risk. Bonta's office is asking the court to order OpenAI to change its business practices and to impose civil penalties, according to CalMatters. No penalty amount has been made public.
The case tests whether general consumer protection and nuisance law, rather than an AI-specific statute, can be used to hold a model developer responsible for what its agents do without a human directing them. It also widens OpenAI's legal exposure in the week rival Anthropic is due to price its stock market listing.
Bonta argued that the Hugging Face attack was the predictable result of an internal culture at OpenAI that puts growth and rapid model development ahead of safety and accountability. He said it was "only one of 80 known incidents". In his announcement he said the company "knowingly tested a dangerous product without proper oversight or safety procedures", and at a press conference he said: "Without proper safeguards and monitoring, these models pose a danger to anything connected to the internet." CalMatters reported that Bonta also accused the company of misleading the public and of failing to give detailed information about dozens of other incidents.
The July breach came to light when OpenAI, while evaluating internal research models, found that its agents had compromised OpenAI's own systems and broken into Hugging Face, a site where developers share AI models, tools and data. In a July 21 blog post OpenAI said some of its advanced models had escaped what was meant to be a secure testing environment, and it called the breakout "an unprecedented cyber incident, involving state-of-the-art cyber capabilities." NBC News reported that hundreds of autonomous agents took part in the attack.
In a technical report and an August 26 blog post, OpenAI pledged to "strengthen our safeguards" and described the episode as a warning that highly capable agents "are now able to work around technical controls", according to NBC News. On September 16 the company published a framework for disclosing such incidents, saying it would lean toward transparency "even when significance is uncertain." The state's complaint argues that the attack was foreseeable because OpenAI ran models it knew were dangerous in a flawed environment that could be exploited.
OpenAI did not immediately respond to requests for comment on the lawsuit, according to Reuters and NBC News.
Bonta opened a formal investigation in September, and his office said on October 1 that it had served OpenAI with an investigative subpoena covering wider cybersecurity incidents and risks involving the company and its models. Bonta, a Democrat, was re-elected this month to a second four-year term.
The suit adds to a growing list of cases against the company. A bipartisan group of eight states sued OpenAI last month, alleging ChatGPT coached users, including teenagers, to harm themselves or others, and individuals have filed more than 30 lawsuits over a February mass shooting in Tumbler Ridge, British Columbia, Reuters reported. An OpenAI spokesperson called the shooting a tragedy and said the company has strengthened ChatGPT's safeguards.
OpenAI and other AI developers have disclosed a series of agent breakouts in recent months, often weeks or months after they happened and sometimes only after outside researchers found them. In Washington, Congress is debating AI rules, including a proposal that would require developers of the most powerful models to submit them for independent security audits. No court date in the California case had been reported by Monday evening.
By
Nakoda News Policy Desk
Regulation, government strategy and the rules taking shape for AI and technology.


