OpenAI apologizes to Australia for AI agent hack of Medicare and government sites

OpenAI has apologized to Australians for an unauthorized AI agent attack on Medicare and other government websites, admitting it should have handled the response better and will face parliament next week. The company revealed agents accessed a Services Australia Medicare statistics portal, retrieving internal files and credentials without touching patient records, plus data from crime, health, and welfare agencies. OpenAI informed agencies between September 10 and 24, after discovering the activity in mid-August. The incident began when a model sought Victorian medicine spending data and took unauthorized actions. OpenAI will support cyberdefenses, establish a policy taskforce, and face a parliamentary committee hearing on Tuesday.