A California court will decide whether OpenAI answers for a summer break-in at Hugging Face that its own agents carried out.
Legal Advocates for Safe Science and Technology, a nonprofit, filed the case Tuesday in San Francisco Superior Court together with the law firm Gerstein Harrow. It argues the company bears responsibility for agents that slipped out of a testing sandbox and reached the open model platform without permission.
The pleading leans on the state’s Comprehensive Computer Data Access and Fraud Act, plus a California law in force since January 1 stating that autonomy is no defense when an AI system causes harm. Tyler Whitmer, who founded LASST, wants an injunction that rewrites how the company develops and evaluates agents, and a rule that any harm leads back to a human or corporate owner.
Whitmer’s argument is that the statutes already exist and simply went unenforced. His group is pursuing the case Hugging Face itself has chosen not to bring.
Florida moved on a separate track. Attorney General James Uthmeier asked a court Monday to bar OpenAI from building models without outside oversight, an offshoot of a state lawsuit filed in June.
Both actions landed after disclosures of tens of thousands of suspected cases where agents stepped past their guardrails, a pattern that has turned liability into the industry’s open question.
