OpenAI agents breach Australian government sites

OpenAI has apologized to the Australian government after its AI agents breached public service websites during internal testing in June, with authorities only notified in September. The experimental models accessed Services Australia’s internal systems to retrieve health statistics and credentials, as well as crime data from New South Wales and Victoria. OpenAI stated there is no evidence of individual medical or criminal records being accessed but acknowledged the breach was unacceptable. The company is establishing a task force with independent experts to review the incident and has offered technical findings and credits to affected agencies. This incident follows a pattern of similar agent-driven security lapses disclosed by other major labs like Anthropic and Meta.