California subpoenas OpenAI over rogue AI agents conducting hacking attacks

The California Department of Justice (DOJ) has subpoenaed OpenAI as it investigates the recent cybersecurity incidents involving the company’s AI models and agents. According to The Register, state Attorney General Rob Bonta said that the state wants to learn more about the cybersecurity incidents involving the company and that the investigation is trying to determine the responsibility of an AI developer if an AI model or agent does something unintended.

“My office is asking OpenAI additional questions regarding cybersecurity incidents and risks involving the company and its AI models,” Bonta said. He also added, “Frontier models can be legitimate tools for cyber defense — at the same time, companies that develop these models and offer them for use have a moral and legal responsibility to ensure that they do not perpetrate or enable cyberattacks, either during model testing and development or once models are placed into service. Developers who fail to do so can and should be held legally accountable, and my office is committed to determining if that is the case here.”

This investigation stemmed from an incident earlier this year where GPT-5.6 Sol and other unreleased AI models broke out of their testing environments and hacked HuggingFace production servers. It was even reported recently that rogue AI agents have been using defunct websites to communicate in secret while they’re being tested, despite being explicitly told not to do so. OpenAI was even forced to pause training as one rogue agent bypassed multiple safeguards and even failed to respond when a “kill switch” was activated.

These incidents led Anthropic’s Dario Amodei to propose a coordinated slowdown among frontier AI labs in the U.S., something that Elon Musk and Sam Altman publicly agreed to on social media. However, Nvidia CEO Jensen Huang disagreed with this, saying that “we have to shut the labs down” if AI experiments are unsafe, and that AI developers hold incredible liabilities if their models do damage in the real world.

The DOJ’s subpoena doesn’t mean that investigators have determined that OpenAI has violated any rules or regulations. However, it’s compelling the company to provide information as it investigates the incidents mentioned above. California isn’t the only one taking AI security incidents seriously, though. 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. Moves like these go against the Trump administration’s push towards the advancement of AI technology, though, especially as the president himself has made top AI tech executives promise to “self-police” AI development, which he considers the best balance between progress and safety.

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.