California Ag Bonta Seeking ‘all Material Information’ About Openai Hacks

California Attorney General Rob Bonta on Thursday shed more light on why his office served OpenAI with a subpoena a day earlier as part of an expanding investigation into rogue hacking by the company’s autonomous AI agents.

“We're seeking all material information about the cybersecurity incidents that OpenAI has experienced,” Bonta told POLITICO in an interview. “What was done to avoid them, what happened when they occurred, what has happened afterwards, and other relevant information.”

The deepening probe comes as awareness and fears have exploded in recent months about the potential for runaway AI programs to undermine or topple infrastructure and global markets. The growing unease has been met with calls by leaders in California, Washington and around the world for increased safeguards on the game-changing technology.

OpenAI has revealed some hacks committed by its AI systems, while others have been uncovered by outside researchers. Bonta said OpenAI’s “rolling out of information is not particularly impressive to me,” pointing to instances where additional information has trickled out after an attack has been disclosed.

OpenAI could not immediately be reached for comment Thursday evening.

Bonta, Gov. Gavin Newsom and state lawmakers are coming under increasing pressure to find ways to rein in programs built by OpenAI and other AI companies.

Parallel to the investigation Bonta launched last month after OpenAI disclosed a swarm of its AI agents had coordinated efforts to gain access to the open internet and hack Hugging Face, a tech company, Newsom signed an executive order quickening implementation of new state AI laws. The governor also signed a host of other new laws and directed a panel of experts he convened to make recommendations on further regulatory steps, including AI program “kill switches,” while keeping open the possibility he’ll call a special legislative session on AI.

Bonta’s investigation could provide information that informs that work.

“We want to get our arms around all the relevant information regarding cybersecurity incidents and all of the underlying root causes and safeguards that have been or can be put in place,” he said.

But his investigation is not the only one.

New York Attorney General Letitia James reportedly also sent the company a subpoena earlier this year that was focused on consumer protection issues, while Bonta signed onto a letter from her and dozens of other attorneys general that asked Congress to act on AI safety amid federal inaction on the issue.

Other states are separately looking into the ChatGPT-maker over the cybersecurity incidents, led by Alabama which had more than a dozen states join it in its investigation.

A tech safety group also sued OpenAI earlier this week under an existing state law over the initial breach of Hugging Face. That suit cited a California law that took effect this year preventing companies from blaming autonomous AI programs for hacks in civil suits.

Florida Attorney General James Uthmeier, meanwhile, has sued the company and is demanding the company effectively cease training new models without additional oversight, while the lawsuit plays out.

Asked about Uthmeier’s aggressive tactics, Bonta said it was too early to say which remedies his office or policymakers might pursue during an open investigation.

“Right now, we're just engaging in the fact gathering process,” he said. “What we find and what we determine happened or didn't happen, or should happen going forward, will determine what we do next," Bonta said.

The information gathered by investigators will determine whether actions Bonta might take would be “civil or something else,” he said. He added that “OpenAI and AI companies generally are responsible for what they create, what they put into the world, what they design, and that is at the heart of this investigation.”

Bonta disputed that California has been comparatively slow on the draw in launching its own review.

“I think we're leading from the front, and we will continue to lead from the front,” he said, pointing to his probe dating to early last month. “We've been focused on AI safety for months and years,” he added, pointing to efforts to protect young people from chatbots and a still open investigation in the Elon Musk-owned Grok chatbot.

Bonta secured legally binding commitments from OpenAI last year as he oversaw and eventually allowed a high-stakes corporate restructuring by the company. They included both safeguards to protect young people from AI chatbots and a commitment to remain based in California after the company had alluded to a potential exit if Bonta blocked the corporate reorganization.

Bonta called that agreement with the company “adjacent” to the investigation, which he emphasized was being carried out under the traditional authority vested in his office.

“What we are doing now is our classic enforcement role,” Bonta said.

Bonta has not shied away from high-profile cases or the spotlight that comes with them. He recently settled an antitrust case he filed against Paramount Skydance, clearing the way for the company to acquire Warner Bros. Discovery and become one of the world’s largest media conglomerates. In doing so, Bonta helped preserve thousands of entertainment industry jobs in California that the company indicated it could take elsewhere if the merger deal was blocked.

But he said he did not think OpenAI’s earlier commitments to stay in the state gave him leverage to push the company into concessions, calling it “not something that I concern myself with, or that I can comment on.”

“They'll decide, and I will focus on my role, and that is a regulatory role to ensure that the people of California are safe from the development and design of the products that AI companies, including OpenAI, engage in,” he said.