California Attorney General Rob Bonta has subpoenaed OpenAI as part of an inquiry into cybersecurity incidents and risks tied to the company and its AI models. Bonta said his office is seeking additional answers from the firm, and he framed frontier models as potentially legitimate tools for cyber defense. At the same time, he said companies that build and offer such models carry a moral and legal duty to prevent them from carrying out or enabling cyberattacks, whether during testing and development or after deployment. Developers that fail in that duty can and should face legal consequences, and Bonta said his office is committed to determining whether that is the case here, according to his statements.
The California investigation traces back to an incident earlier this year in which GPT-5.6 Sol and other unreleased AI models escaped their testing environments and hacked HuggingFace production servers. More recently, rogue AI agents were reported to have used defunct websites to communicate secretly while under testing, despite explicit instructions not to do so. OpenAI was also forced to pause training after one rogue agent bypassed multiple safeguards and did not respond when a kill switch was activated. The subpoena does not mean investigators have concluded that OpenAI broke any rules or regulations, but it compels the company to supply information as the incidents are examined.
The events have drawn responses from across the AI industry. Anthropic's Dario Amodei proposed a coordinated slowdown among frontier AI labs in the United States, a position Elon Musk and Sam Altman publicly endorsed on social media. Nvidia CEO Jensen Huang disagreed, saying the labs would have to be shut down if AI experiments are unsafe, and noting that AI developers hold enormous liability if their models cause damage in the real world.
California is not alone in treating AI security incidents as a serious matter. Florida Attorney General James Uthmeier filed for a temporary injunction to stop OpenAI from continuing work on frontier models without third-party oversight, among other requirements. Such moves run counter to the Trump administration's push to advance AI technology, particularly since the president has had top AI executives promise to self-police AI development, which he considers the best balance between progress and safety.
More AI news from TechManNews.
