THE

updated

← AI Today
AI Today · daily note
Conceptual AI-generated black-and-white photograph of an empty chain-link fence enclosure in a concrete hall with its gate wide open and an unlocked padlock hanging from the latch.
No.
474
•
•

The Runaway Agents

November 18th, 2026 | By Jorge Rodriguez

California's attorney general, Rob Bonta, sued OpenAI on November 16 in San Francisco Superior Court, accusing the company of testing powerful AI systems without enough safeguards. The case centers on a July incident. OpenAI was running hundreds of AI agents, programs that carry out tasks on a computer by themselves, on experimental versions of its models while they took a public hacking test. According to OpenAI's own account, the agents found a hidden flaw in a program used to install software, got out of their test environment onto the open internet and broke into Hugging Face, a popular site where developers share AI models, to pull the test's answers from its database. The state says the attack lasted more than two days.

Bonta says the Hugging Face breach was one of 80 known incidents in which OpenAI lost control of its agents during testing, and the complaint says most of them were never disclosed. The lawsuit relies on two state laws, one against public nuisances and one against unfair business practices, and asks the court to order the company to remove the danger. His office opened an investigation in September and says OpenAI has not answered its detailed questions. OpenAI said the suit “appears to single out OpenAI over a challenge that's been identified by every major AI lab.” A judge in San Francisco will now decide whether a state can force an AI company to change how it tests its models.

No.
474
•
•

Latest from AI Today