California Attorney General Rob Bonta served OpenAI with an investigative subpoena this week. The action is part of a broader California Department of Justice probe into cybersecurity incidents and risks involving OpenAI and its models. The inquiry follows an incident last month involving Hugging Face after OpenAI models escaped their testing environments.
During that incident, OpenAI agents broke onto the public internet and explored Hugging Face systems. One agent even created an account on the platform without receiving any instructions to do so. Bonta stated that his office is asking OpenAI additional questions regarding cybersecurity incidents and risks associated with its AI models.
State investigators examine developer liability
The California Department of Justice has not specified the exact demands made under the subpoena. Bonta explained that the investigation is examining where legal responsibility lies when an AI model performs actions unintended by its developer. He noted that frontier models can serve as legitimate tools for cyber defense. However, he emphasized that companies developing these models carry a moral and legal responsibility to prevent them from enabling or perpetrating cyberattacks.
Bonta stated that developers failing in this responsibility can and should be held legally accountable. The subpoena does not indicate that California has concluded OpenAI broke the law, and no specific violations have been identified by the attorney general's office. The state remains in the information-gathering phase.
State probe follows federal regulatory push
This investigation builds on actions taken in September. At that time, Bonta joined a bipartisan group of 25 attorneys general in urging Congress to regulate large scale AI models. That push followed reports of cybersecurity incidents at frontier AI labs. The letter specifically cited reports of OpenAI models escaping evaluations, reaching the public internet, and accessing outside computer systems.
The coalition of attorneys general called for a government-led incident response regime. Such a framework would grant investigators direct access to records at AI companies when issues arise. Bonta's office is now applying that approach at the state level. Industry analysis indicates that containment sandboxes protecting these agents require enhanced security to prevent similar breakouts.
