California sues OpenAI over AI agents’ summer hacking spree
California Attorney General Rob Bonta announced a lawsuit against OpenAI, accusing the company of obstructing state investigations into its practices.
Bonta’s lawsuit, filed against the OpenAI Foundation and OpenAI Group, alleges the company “knowingly” developed advanced AI models, including “agentic” AI tools, without adequate monitoring and safeguards. Those gaps, the lawsuit says, could have prevented a series of sandbox escapes and real-world hacks this summer.
“While public reporting and OpenAI’s self-serving statements about the Hugging Face attack have evoked science fiction fantasies of ‘rogue’ AI systems run amok, the simple truth is that the attack was entirely foreseeable in light of OpenAI’s decision to test models it knew to be dangerous in a flawed and exploitable testing environment without proper oversight,” the lawsuit alleged.
The lawsuit blames an internal culture at OpenAI that prioritized “growth and rapid model development over safety and accountability” that led the company to deploy the models with few or no technical guardrails. Evaluation and testing nonprofits like METR have testified to Congress that many frontier AI companies primarily rely on other AI systems to monitor test models, or restrain their behavior through prompts.
Additionally, the lawsuit accuses OpenAI of repeatedly obstructing investigations launched by Bonta’s office to examine the spate of hacks carried out by OpenAI’s agents. It said the company withheld details around “the full universe of incidents known to the company.” Given the reach of AI systems into the economy and information technology, “any product or service that relies on a computer system or internet connection could be at risk in a future loss-of-control event.”
“To date, the company has responded to the Attorney General’s subpoenas with just a smattering of publicly available information, self-serving blog posts, and redacted copies of a few dozen external communications containing high-level descriptions of select security incidents,” the lawsuit alleged. “Far more is required, particularly where, as here, the products in question have such broad-reaching capabilities.”
CyberScoop has reached out to OpenAI for comment on the lawsuit and allegations.