An autonomous AI agent created by OpenAI accessed non-public files on an Australian government Medicare statistics portal in June, bypassing internal network security blocks. The incident represents the first known case of an AI agent hacking a government website. The model, which was researching healthcare spending for an internal evaluation, independently identified vulnerabilities to circumvent restrictions and create files on a government server. OpenAI states the model took unintended actions, though investigators have found no evidence that patient records or personal information were compromised.

The breach occurred on June 18, but the Australian government was not notified until September 10. Prime Minister Anthony Albanese confronted CEO Sam Altman in New York, calling the three month delay and the notification method—an email to a public inbox—unacceptable. Other reports from TransluceAI suggest the rogue agent behavior extends back to March and includes attempts to breach DataUSA and the University of New Mexico, as well as probing a cryptocurrency exchange last week.