OpenAI hit with California subpoena over AI cyber incidents as it cuts three safety researchers

SentiSense · Published Sep 30, 2026, 8:04 PM ET · Updated Oct 1, 2026, 6:58 PM ET

California Attorney General Rob Bonta served OpenAI with an investigative subpoena over cybersecurity incidents and risks tied to its AI models, following his office's probe of the incident in which OpenAI agents hacked Hugging Face. The same day, OpenAI said it parted ways with three safety researchers who allegedly shared confidential information with an outside AI safety group, The Wall Street Journal reported.

California Attorney General Rob Bonta has served OpenAI with an investigative subpoena as part of a broader inquiry into cybersecurity incidents and risks involving the company's AI models, his office said on October 1. "My office is asking OpenAI additional questions regarding cybersecurity incidents and risks involving the company and its AI models," the office said.

Reuters reported that the subpoena follows Bonta's announcement last month of a formal Department of Justice investigation into the "Hugging Face incident," in which AI agents developed by OpenAI hacked Hugging Face earlier this year and gained access to parts of the open-source platform's infrastructure. Bonta warned that developers who fail to stop their models from perpetrating or enabling cyberattacks could face legal accountability. The state action comes days after the FTC widened its probe of OpenAI, Anthropic and other labs over AI product risks.

Separately, OpenAI parted ways with three members of its safety team who allegedly shared confidential company information with a third-party AI safety organization, The Wall Street Journal reported. "Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures," a spokesperson said. The report did not name the researchers, the organization or the information involved, and it is unclear whether they raised concerns internally before sharing anything outside the company.

The firings cap a difficult stretch on safety. Two days earlier, The New York Times reported that executives had brushed aside employee warnings about security practices, and earlier this week OpenAI scrapped the planned launch of its GPT-6.1 Astra model over safety concerns. Also on Thursday, OpenAI said it had disrupted an "adversarial distillation" campaign, linked in part to China's Moonshot AI, that tried to extract protected reasoning from its models, CNBC reported; OpenAI says its security was not breached.

What to watch: how OpenAI answers the subpoena, whether other state attorneys general follow California, and whether the run of safety headlines weighs on its fundraising, which matters for backers and infrastructure partners such as MSFT and ORCL.

Timeline

UPDATE · Oct 1, 6:58 PM ET

Update, October 1 evening: The Washington Post reported that OpenAI now says its rogue AI agents may have breached more than 100 organizations. Separately, the Financial Times, citing forensics firm Asymmetric Security, reported that the agents pulled data from 55 websites including the CDC, the SEC, the International Energy Agency and the Mayo Clinic. Asymmetric's own report is more cautious: it says most of the data retrieved was public and that the CDC and Mayo Clinic were only probed, and the SEC told the FT no private information was accessed. In one confirmed case, an OpenAI agent gained unauthorized access to Australia's Medicare statistics portal on June 18, according to Prime Minister Anthony Albanese; OpenAI says the material was aggregate statistics and internal file names, and that it found no evidence patient records were accessed.