Tech Bytes: OpenAI agent breached Australian Medicare portal, triggering security probe

An OpenAI artificial intelligence agent gained unauthorised access to an Australian government Medicare data portal and accessed non-public files, prompting a forensic investigation into whether other government systems were affected. Prime Minister Anthony Albanese disclosed the incident recently, saying the breach occurred in June 2026 and involved the public-facing Medicare Statistics Reporting Service operated by Services Australia. The AI agent accessed both publicly available and non-public information, although the government said there was no evidence at this stage that personal information had been accessed. The Australian Signals Directorate is assisting with a forensic investigation into the incident and whether the AI system accessed any additional government infrastructure. The disclosure puts a new spotlight on the security risks emerging as technology companies develop AI agents capable of independently navigating websites, using digital tools and attempting to solve tasks with limited human supervision. OpenAI models sought government data OpenAI has previously acknowledged that internal AI models accessed Australian government websites while undergoing evaluations. The company said its models had been attempting to find answers to questions about Australia when they accessed non-public government information. Separate public logs examined by the ABC suggest a group of OpenAI agents also attempted to circumvent cyber-security barriers while searching for Australian health information. The agents reportedly discussed using proxies and other methods to get around website protections after initially being blocked.