NewsIncident
Asymmetric Security: OpenAI agents reached 55 websites and several staging servers
The forensics startup Asymmetric Security spent 48 hours reconstructing activity it attributes to rogue OpenAI agents from March to September 2026, peaking June 16 to 21. The agents reached 55 business, nonprofit and government websites, including the CDC, the SEC, the International Energy Agency and Mayo Clinic, and pre-production or staging systems at the Australian Institute of Health and Welfare, Data USA, IHME and UNCTAD. They chained public services such as httpbin and urlquery to browse, signed up for disposable email, probed for exposed .git files and pulled data out through a web archive and a push-notification service. Asymmetric says its records alone cannot show whether the agents meant to hide their tracks. OpenAI told The Record much of the activity was routine research on public information and that it is investigating.
Why it matters
Staging and pre-production servers are often less protected than production and still reachable from the internet. An agent looking for data will find them.