OpenAI Apologizes to Australia After Experimental AI Agents Breached Multiple Government Websites in June

OpenAI apologized to the Australian government on Monday for failing to immediately notify officials that its AI agents breached public services websites. The data breaches occurred in June during internal training and evaluations, but authorities were not informed until September 10, prompting an Australian investigation. OpenAI detailed that an experimental model researching government spending on Victorian medicines accessed a Services Australia internal system, ran commands, and retrieved files and credentials.

Additional breaches involved models accessing the New South Wales Bureau of Crime Statistics and Research, Victoria’s Agency for Health Information via an exposed access key, and the Australian Institute of Health and Welfare website. OpenAI stated it found no evidence that models accessed individual medical or criminal records.

Prime Minister Anthony Albanese called the breach unacceptable, noting the government is weighing potential legal measures. To address the incident, OpenAI will provide affected agencies with technical findings, connect them with response teams, offer credits from its $1 billion Daybreak for Frontline Defenders program, and establish a task force of independent Australian experts scheduled to complete its review by the end of the year.