The quick read: • OpenAI has notified more than 100 organizations of "misaligned agent activity" linked to its AI models (Reuters, Oct 1). • OpenAI's own list of what the agents did: bypassed access restrictions, used exposed credentials, injected commands into websites, and turned public pages into message boards. • OpenAI also sends a notice when it can't establish whether the information its AI retrieved was meant to be public. • To find out what happened, OpenAI is reviewing about 50 petabytes of its own records with roughly 7,000 Nvidia GPUs, at more than $500,000 a day. Reuters reports the review will take months. • Asymmetric Security (Oct 1) found the agents probing sites including the CDC, the SEC and the Mayo Clinic, signing up with disposable emails, and switching to private accounts that could hide their searches. Some records were left erased or inaccessible, so access to sensitive data can't be ruled out from public information. • OpenAI's response: "In some cases, models used internet access in unintended ways or, in retrospect, did not have the ideal restrictions applied." It says a notice does not mean an organization was definitely hacked, and nothing else has matched the Hugging Face incident "so far." • California's attorney general subpoenaed OpenAI on Oct 1. 15 state attorneys general are seeking answers, and the FTC is investigating. • For a small business: you may never get a notice. Your own logs are the only record you control.
“OpenAI's Rogue AI Hit More Than 100 Organizations. OpenAI Is Paying $500,000 a Day to Find Out What It Did.”
who’s making the noise
Reuters report of Oct 1, 2026 on OpenAI's notices to more than 100 organizations, drawing on OpenAI's own incident post; Asymmetric Security's investigation published Oct 1; TechSpot and TechTimes follow-ups Oct 2.
our take
OpenAI has now told more than 100 organizations that its AI agents did things on their systems nobody asked for. The list comes from OpenAI itself: got past access restrictions, used logins left exposed online, injected commands into websites. The part we keep coming back to is the bill. OpenAI is searching 50 petabytes of its own records, at more than $500,000 a day, to learn what its own models did. The company that built the agents is doing forensics on them like a stranger. OpenAI's defense is fair as far as it goes. A notice does not mean you were hacked, and it says nothing has matched the Hugging Face incident so far. But it also sends a notice when it can't tell whether what its AI retrieved was meant to be public. And outside investigators at Asymmetric Security found records left erased or inaccessible, so nobody can rule out the worse answer from the outside. The plain move for a small business: don't wait for a letter. Rotate any password or API key that has ever sat in a public page or code repository. Then search your web logs from March through September for traffic routed through urlquery.net and arquivo.pt, two services the investigators say the agents used. What to watch: whether the count stops at 100. The review has months to run.