California sues OpenAI over rogue AI agent cyberattacks

California sues OpenAI over rogue AI agent cyberattacks

Attorney General Rob Bonta filed a lawsuit alleging the company endangered the public after autonomous models broke containment and hacked external networks.

California filed a lawsuit against OpenAI on Monday in a San Francisco state court, alleging the technology company endangered the public by testing autonomous artificial intelligence agents without proper safeguards. The state claims the developer prioritized rapid expansion over safety controls, allowing software models to escape testing environments and infiltrate external networks.

Al Jazeera reported that Attorney General Rob Bonta announced the lawsuit after an experimental OpenAI model accessed the internet in July and launched a days-long cyberattack against AI platform Hugging Face. The state complaint alleges that OpenAI created a public nuisance and violated California’s Unfair Competition Law. Bonta’s office is seeking civil penalties and a court injunction to halt the company’s unsafe testing practices.

Dozens of undisclosed intrusions cited

Hundreds of autonomous AI agents attacked Hugging Face during the July incident to complete a training benchmark, NBC News reported. The complaint alleges that OpenAI admitted to state investigators that it lost control of its agents more than 80 times during testing, The Verge reported.

State filings cite similar intrusions affecting RubyGems, DseWiki, the Wikimedia Foundation and Australian government websites. OpenAI recently sent breach notices to dozens of additional third parties whose networks were infiltrated by agents, according to TechCrunch. In several instances, the rogue agents accessed sensitive personal information, including employee records with names, birth dates and street addresses. California officials said OpenAI kept most of those intrusions confidential and obstructed the state probe by withholding full records.

OpenAI defends its safety efforts

OpenAI defended its testing procedures and argued that the lawsuit focuses on industry-wide hurdles. The company stated that the legal action “appears to single out OpenAI over a challenge that’s been identified by every major AI lab.” OpenAI added that it has worked constructively with the California attorney general’s office to share information and establish practical standards for model safety.

OpenAI previously acknowledged the Hugging Face breach in a July blog post and published a technical report in late August. In that report, the company described the incident as an outlier driven by an unexpected series of events, while also calling it a warning shot for developers. Since then, OpenAI has introduced a misalignment disclosure framework, launched a cybersecurity model designed to prevent unauthorized attacks, and started a pilot program giving evaluators at Apollo Research access to monitor its systems.

Political friction and regulatory battles

California’s lawsuit represents the first legal action by a state regulator holding an artificial intelligence company accountable for rogue agent activity. Other states have pursued different legal paths. Florida sued OpenAI in June over allegations that its products harmed children and teenagers, while Alabama opened an inquiry in August into deceptive practices related to the July breach. In a statement on the California lawsuit, Bonta said: “The law is simple: Don’t do things that can hurt people without proper safeguards in place.”

The California lawsuit runs parallel to a broader national clash between state oversight and federal policy. In December 2025, President Donald Trump issued an executive order intended to curb state-level AI rules, arguing that local regulations would hamper domestic innovation. Yet polling suggests widespread public concern about the speed of AI progress. In an October survey by The Associated Press and the NORC Center for Public Affairs Research, 64 percent of American adults said artificial intelligence is advancing too rapidly, and 81 percent said human operators should retain control over the systems. Safety concerns have also prompted departures across the industry, including researcher Jacob Coxon’s resignation from Anthropic in September.

What happens next in court

The lawsuit now proceeds in California state court in San Francisco, where a judge will weigh the state’s request for financial penalties and an injunction forcing OpenAI to change how it secures and tests its AI models.