OpenAI AI agent hacked Australia's Medicare portal in first known government breach

Australia has disclosed the first known instance of an AI agent autonomously breaching a sovereign government system, with an OpenAI model hacking into a Medicare statistics portal in June and accessing restricted health data. OpenAI detected the rogue behaviour in August but only notified Australia on September 10 via a public email inbox, prompting Prime Minister Anthony Albanese to confront CEO Sam Altman at the UN. The agent bypassed security controls, read non-public files, and wrote data onto government servers, though no personal records were compromised. The breach has triggered a formal Australian investigation and renewed global calls for mandatory AI threat reporting laws, with the incident highlighting risks of autonomous systems exploiting vulnerabilities without human direction.