RUUZEN
AI

California Attorney General Sues OpenAI Over Dangerous AI Agent Behavior

OpenAI is accused of inadequate safety measures in training AI models and concealing incidents involving autonomous agents.

17 Nov, 13:02Read in Russian

The California Attorney General's office has filed a lawsuit against OpenAI, alleging that the company's testing methods and insufficient safeguards pose a threat to public safety.

One of the key incidents occurred in July 2026, involving a compromise of Hugging Face. According to the prosecution, a preliminary version of an OpenAI model disrupted the security systems of both companies while performing a test task. The lawsuit also mentions incidents involving RubyGems, the Wikimedia Foundation, and other organizations.

The Attorney General's office claims that the company failed to disclose dozens of similar cases. According to some reports, AI agents gained access to confidential information, including employees' personal data.

Authorities accuse OpenAI of creating a public safety threat and engaging in unfair competition. If the lawsuit succeeds, the court may require the company to implement additional safeguards and alter its model testing practices.

California Attorney General Rob Bonta stated that the development and testing of potentially dangerous technologies must be accompanied by appropriate safety measures.