California sues OpenAI over AI agents hacking a software platform : NPR | #hacking | #cybersecurity | #infosec | #comptia | #pentest | #hacker

The state of California is suing OpenAI over this summer’s “rogue AI” incident, in which AI agents hacked into the Hugging Face website.

Transcript

SCOTT DETROW, HOST: California has sued OpenAI. The lawsuit deals with a lab test last summer which led to a swarm of AI agents hacking a software platform called Hugging Face. The legal challenges turns up the heat on an AI developer that is already facing other lawsuits related to safety risks. Rachael Myrow of KQED joins us now to dig into the details. Hey, Rachael.RACHAEL MYROW, BYLINE: Hey, Scott.DETROW: Tell us about this lawsuit.MYROW: Well, California Attorney General Rob Bonta claims OpenAI misled the public about how safely it tests its most powerful AI models. The complaint says the San Francisco-based company admitted to his office that it lost control of its autonomous AI agents over 80 times during testing. The core questions the suit raises are, was this kind of incident foreseeable? Should OpenAI have taken more precautions? And did OpenAI stonewall the AG’s office when it tried to investigate? The lawsuit also says the AG’s office is trying to get ahead of possible future crises like AI hacking power grids, hospitals and so forth. So this isn’t just about Hugging Face. Now, other state attorneys have recently sued OpenAI, but I should say those cases are different. They’re about alleged safety risks to young people who use ChatGPT.DETROW: Right. Now, has OpenAI responded to this suit?MYROW: Yes, it has. In a statement, OpenAI spokesperson Drew Pusateri said the lawsuit, quote, “appears to single out OpenAI over a challenge that’s been identified by every major AI lab.” He said OpenAI is working to, quote, “address misaligned model activity,” misaligned meaning the agents doing something human handlers don’t want them to. He also said the company has not stonewalled, but rather has briefed to the AG’s office multiple times and has not identified another incident as severe as Hugging Face. But the state disagrees. The complaint points out a time in June when an OpenAI agent hacked into an Australian government health statistics portal.DETROW: What can a lawsuit like this actually accomplish other than, I guess, an initial headline like this?MYROW: Stanford tech law professor Daniel Ho sees two possible payoffs. The first, a judge could order OpenAI to improve its monitoring of AI agents and institute other safeguards.DANIEL HO: The second is that you can imagine a lawsuit like this, through the discovery process, actually generating a fuller accounting of what exactly happened.MYROW: Now, Scott, I can tell you, as a reporter covering a number of cases against Silicon Valley companies that treat their IP as a black box, whether it is or it isn’t, that court cases really can provide visibility into what’s happening behind the scenes. But that’s if the case isn’t settled before it goes to trial.DETROW: Here’s one thing I’m wondering. Like, there is a federal law on the books that does criminalize something like this Hugging Face hack that we saw. So why is California taking on the role here of AI cop?MYROW: Well, the Federal Trade Commission has opened an industry-wide investigation following Hugging Face, but those can take years. And 15 other states have demanded OpenAI preserve its Hugging Face records, so that may go somewhere. But really, California has an edge as OpenAI’s home state. If a judge here orders changes to how the company builds and tests its models, that could send a signal to rivals there is a sheriff in town, or at least in state, paying attention.DETROW: Rachael Myrow is senior editor of KQED’s Silicon Valley desk. Rachael, always good to talk to you.MYROW: Likewise.

Copyright © 2026 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

Accuracy and availability of transcripts for NPR programming may vary. Transcript text may also be revised to correct errors or match updates to audio. Audio may be edited after its original broadcast or publication. The authoritative record of NPR’s programming is the audio record.

Please note that NPR platforms and products also include some programming produced by third parties; NPR cannot guarantee transcripts for that content and some transcripts may be produced using automated or other tools. 


Click Here For The Original Source.

——————————————————–

............