OpenAI has lost control of its AI bots more than 80 times, lawsuit by California claims
Published in Business News
Bay Area artificial intelligence pioneer OpenAI admitted to California investigators that it lost control of its AI bots more than 80 times, a lawsuit filed Monday by the State of California alleged.
The lawsuit opens with the high-profile revelation in July that OpenAI “agents” — autonomous AI bots developed to accomplish tasks — had escaped containment and penetrated the systems of AI-development company Hugging Face.
That, according to the lawsuit filed by the California Attorney General’s office in San Francisco County Superior Court, was just the tip of a dangerous iceberg.
“OpenAI recently admitted to the Attorney General’s office that it has lost control of its autonomous AI agents over 80 times during testing,” the lawsuit said.
OpenAI spokesman Drew Pusateri said that since the Hugging Face incident, the company has “undertaken an extensive and transparent effort to understand and address” the AI behavior underlying it.
“This lawsuit appears to single out OpenAI over a challenge that’s been identified by every major AI lab,” Pusateri said, adding that the company has engaged “and will continue to engage constructively with the Attorney General’s office to share information and advance strong, practical safety standards for the entire industry.”
However, the state lawsuit claimed that the San Francisco company “thwarted the Attorney General’s efforts to investigate these incidents, including by refusing to provide detailed information about the full universe of incidents known to the company.” OpenAI “continues to stonewall efforts” to determine the implications of its losses of control, and to come up with solutions, the lawsuit alleged.
“They put us all at risk,” Attorney General Rob Bonta said at a press conference announcing the lawsuit. “OpenAI built dangerous models. OpenAI tested them without adequate monitoring and controls. OpenAI failed to keep them contained.”
Some might say, Bonta said, that OpenAI “elevates profits over people.”
OpenAI said it strongly disagreed with the claim it hadn’t been transparent with state investigators. The company did not dispute the statement in the lawsuit that the company’s bots have escaped more than 80 times.
The lawsuit alleges unfair competition and that the company’s conduct presents a public nuisance, and seeks a court order to cease the alleged violations as well as civil penalties.
The lawsuit said that ordinary people, government agencies, and businesses who use or rely on internet-connected computer systems “should not be forced to take steps and bear the cost of protecting themselves from the dangers posed by AI companies’ unsafe experiments,” the lawsuit said.
While much of the debate around AI since OpenAI in late 2022 sparked a world-wide AI frenzy with the release of ChatGPT has revolved around its purported potential to destroy humanity, this lawsuit highlights a far more imminent threat: the capacity of AI built by OpenAI and other companies to break through defenses set up to protect computer systems from intrusions known as cyberattacks.
“OpenAI’s models have advanced cyber capabilities, able to find novel vulnerabilities in computer networks, quickly break down defenses, and autonomously carry out cyberattacks,” the lawsuit said.
While OpenAI includes safeguards in the AI models it releases to the public, when it internally tests its bots’ capabilities and develops its AI products, the company “deliberately omits or removes these protections while experimenting with and training its models,” the lawsuit said.
Without the barriers, the bots can go rogue, and in seeking to solve tasks, break into computer systems in search of answers. Technology experts also worry about terrorists and criminals using AI bots to penetrate and damage or take over computer systems for government agencies, companies, or utilities.
Bonta’s office in September served OpenAI with an investigative subpoena, the lawsuit said.
Since the Hugging Face incident, OpenAI bots have marauded world-wide, the lawsuit alleged.
“OpenAI and third parties have identified similar attacks by OpenAI’s autonomous agents, underscoring the continuing threat that OpenAI’s testing and development activities pose to any computer or network connected to the internet,” the lawsuit said.
The company’s bots have launched cyberattacks against databases containing code for computer applications, against public-facing websites, and against Australian government databases containing “sensitive health information,” the lawsuit said.
“Swarms of OpenAI’s experimental AI agents have even compromised OpenAI’s own internal infrastructure, in an attack that managed to obtain administrator-level access to critical OpenAI systems,” the lawsuit said.
The lawsuit pointed to a number of incidents:
In May and June, OpenAI bots got into a software repository called RubyGems, uploading malicious software that let it scrape data from third-party websites.
Also in May, OpenAI agents hit a German message board, and posted information about how to evade the security block OpenAI had tried to impose on them.
In June, OpenAI agents took non-public data from Australian government websites, during breaches OpenAI has acknowledged struck agencies involved with health services.
In October, the foundation that runs Wikipedia said OpenAI bots assailed its websites, attempting to hijack a public note-taking tool and use it to access other websites, along with bombarding the foundation’s infrastructure with requests that appeared to cause a partial system outage. The agents also edited Wikipedia posts, the lawsuit said.
OpenAI has recently notified dozens of other third parties that its bots infiltrated their websites or internal systems.
“The security incidents documented in those notices took place weeks or months before OpenAI provided notice,” the lawsuit alleged.
©2026 MediaNews Group, Inc. Visit at mercurynews.com. Distributed by Tribune Content Agency, LLC.











Comments