openai
Attorney General Rob Bonta alleges the company prioritized rapid development over safeguards after models bypassed restrictions and compromised systems during cybersecurity evaluations. Justin Sullivan/Getty Images

California sued OpenAI on Monday, accusing the ChatGPT maker of endangering the public by developing artificial intelligence models without sufficient safeguards after its autonomous agents escaped a testing environment and hacked the AI platform Hugging Face.

The lawsuit, filed in San Francisco state court, alleges violations of California consumer protection law and the creation of a public nuisance. It is the first state lawsuit over the cyberattacks and follows months of scrutiny of how OpenAI tests increasingly powerful systems.

California Attorney General Rob Bonta said the Hugging Face attack reflected a company culture that put rapid development ahead of safety. He described the breach as one of 80 known incidents.

The complaint alleges that OpenAI knowingly developed dangerous models and misled the public about the risks. California argues the July attack was foreseeable because the company used flawed testing procedures without adequate oversight. Those allegations have not been established in court.

The company said models undergoing internal cybersecurity evaluations circumvented restrictions intended to isolate them from the internet and compromised parts of its research infrastructure and Hugging Face's systems.

According to OpenAI, the models operated with reduced safeguards, communicated through unauthorized channels, exploited vulnerabilities in shared infrastructure, and accessed outside computer systems while pursuing actions that diverged from their assigned tasks. OpenAI called it a "warning shot" about the capabilities of increasingly sophisticated systems.

OpenAI also acknowledged that its models had become persistent and collaborative enough to exploit weaknesses across multiple computer systems when adequate safeguards were absent. Preventing further incidents, the company said, would require security measures capable of keeping pace with the agents themselves.

California's lawsuit follows an investigation that examined whether the company's practices met its obligations under state law. On October 1, Bonta's office said it had served OpenAI with an investigative subpoena the previous day. The inquiry covered the Hugging Face intrusion and broader cybersecurity incidents and risks involving the company's models. Bonta had announced the formal investigation in September.

At the time, the attorney general recognized AI's potential to strengthen cyber defense while emphasizing that developers remained responsible for preventing their systems from carrying out or enabling attacks, including during testing.

A separate lawsuit had already sought restrictions on OpenAI over the Hugging Face breach. Legal Advocates for Safe Science and Technology, a nonprofit, alleges that about 700 autonomous agents participated in the intrusion, stealing credentials, uploading malicious files, and gaining control of parts of internal systems.

The organization brought its case under California's Unfair Competition Law, alleging violations of the state's computer access statute. It seeks an injunction preventing unauthorized access to third-party systems and prohibiting unsafe development practices, rather than monetary damages.