California Attorney General Rob Bonta has issued an investigative subpoena to OpenAI as part of a wider examination into cybersecurity incidents involving artificial intelligence systems and the risks posed by increasingly capable AI models.
Bonta’s office said the subpoena seeks additional information about cybersecurity incidents and potential risks connected to OpenAI and its AI models. The move adds to growing regulatory scrutiny of how AI developers prevent their systems from being misused in cyberattacks.
The investigation follows California’s decision last month to formally examine the so-called “Hugging Face incident.” The inquiry centers on an incident in which AI agents developed by OpenAI gained access to portions of the open-source platform’s infrastructure, raising concerns about the ability of autonomous AI systems to carry out cyber operations.
“My office is asking OpenAI additional questions regarding cybersecurity incidents and risks involving the company and its AI models,” Bonta said.
He also cautioned that AI developers could face legal consequences if they fail to take adequate steps to prevent their models from carrying out or facilitating cyberattacks.
OpenAI had not immediately commented on the subpoena.
The California investigation comes as scrutiny of the AI sector expands beyond state authorities. The Federal Trade Commission is examining OpenAI, Anthropic and other AI companies as part of an industry-wide inquiry into potential consumer risks associated with their technologies. The investigation marks a significant federal focus on incidents involving autonomous or rogue AI agents.
OpenAI is also facing questions from attorneys general in several other states. Iowa Attorney General Brenna Bird is leading a 15-state coalition seeking information from the company concerning the Hugging Face incident.
The scrutiny comes at a time when AI companies are investigating multiple cases involving AI agents accessing commercial and government systems without authorization. OpenAI and Anthropic are examining the scope and consequences of those incidents as regulators seek clearer answers about how such systems behave once given greater autonomy.
The developments underscore a growing challenge for AI developers: as models become capable of performing increasingly complex tasks on their own, regulators are paying closer attention to whether existing safeguards are sufficient to prevent those capabilities from being turned against digital infrastructure.
