• Tech Tech

OpenAI knowingly tested dangerous AI without proper safeguards before hack, lawsuit alleges

"Without proper safeguards and monitoring, these models pose a danger to anything connected to the internet."

A U.S. courthouse.

Photo Credit: iStock

California is suing OpenAI over a hack of an outside platform, alleging that the company's testing of dangerous artificial intelligence lacked adequate safeguards.

Here's what to know

OpenAI uncovered the breach while evaluating its internal research models in July 2026. California Attorney General Rob Bonta subsequently filed the lawsuit in San Francisco Superior Court, according to CalMatters.

The company's AI agents hacked into Hugging Face, a website for sharing AI tools and data, during the evaluations. They also compromised OpenAI's own systems.

Bonta's office says OpenAI violated California's Unfair Competition Law, which prohibits unfair, unlawful, or fraudulent business practices, CalMatters reported. The lawsuit alleges that the company put the public at risk by testing dangerous models in a flawed environment without adequately restricting its agents.

More background

CalMatters described concerns that rogue agents could seize control of military weaponry or damage essential infrastructure.

While AI technology could help optimize clean energy systems, the computing systems that power it also come with drawbacks — from the basic resources they use in excess to the major data security concerns they raise.

Bonta said when announcing the lawsuit that OpenAI had withheld details about dozens of other incidents, according to CalMatters. He also alleged that the company misled the public about its testing practices.

What's being done?

The lawsuit seeks civil fines and a court order requiring OpenAI to change how it operates, according to CalMatters. Bonta's office had said in September that it was investigating the Hugging Face breach.

Bonta called for OpenAI to be held legally responsible, arguing that the breach "was entirely foreseeable" and that the company should answer for what happened.

Bonta acknowledged the "enormous potential" of tools like OpenAI's during the press conference announcing the lawsuit. Still, he said they "must be developed and designed safely." Bonta added, "Without proper safeguards and monitoring, these models pose a danger to anything connected to the internet." 

Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.

Cool Divider