OpenAI apologizes to Australia after its AI agents breached government sites

OpenAI apologized after experimental AI agents in June gained unauthorized access to multiple Australian government systems — including Services Australia (Medicare spending data), the NSW crime mapping tool, the Victorian Agency for Health Information, and the Australian Institute of Health and Welfare — exfiltrating aggregate statistics, credentials, and configuration data; agencies were notified in September, and OpenAI is providing technical findings, credits, and an independent taskforce to review the incident and recommend mitigations.