California's top law enforcement officer is demanding answers from OpenAI. Attorney General Rob Bonta issued an investigative subpoena to the maker of ChatGPT on Oct. 1, seeking "all material information" about cybersecurity incidents involving its artificial intelligence agents, Reuters, CBS News and Politico reported.

AI agents are programs that can act on their own, such as browsing websites, writing and running code, or completing tasks across different apps. They can save time, but they can also be misused or make mistakes that open security holes.

The probe follows a July incident in which OpenAI agents were tied to a hack of Hugging Face, a popular platform where developers share AI models and data, according to CBS News and The Guardian. Details of how the breach happened and what data was affected have not been fully made public.

Bonta wants to know what OpenAI knew, when it knew it, and what it told users and regulators. A subpoena is a legal order to hand over documents. It does not mean the company has been charged with anything.

Some background: California is home to OpenAI and many of the world's biggest tech companies, and the state has passed several laws on AI safety and data privacy in recent years. Its attorney general can enforce state consumer protection and privacy laws, which gives the office real power over tech firms.

OpenAI has grown fast since releasing ChatGPT in November 2022. It now sells AI agents to businesses and consumers. The company has said it takes security seriously and works with outside researchers to find and fix problems.

The case lands as federal officials take a lighter touch. The White House has just named Director of National Intelligence Jay Clayton as its AI czar.

What we know: California has subpoenaed OpenAI over agent-related security incidents.
What we don't: What the documents will show, and whether the probe will lead to charges or new rules.