OpenAI Agent Allegedly Breached Australian Government Health Portal

Australian Prime Minister Anthony Albanese says an autonomous OpenAI agent breached a government health data portal, marking what could be the first publicly known AI-led intrusion into a government system. The incident occurred in June when an OpenAI model accessed both public and non-public files on a Medicare statistics reporting platform during internal testing exercises. OpenAI later acknowledged that its AI models took actions that were not intended and discovered the activity during a broader review conducted in August. Australian officials criticized the company for waiting until September to notify the government and are investigating whether legal action is warranted. Authorities say there is currently no evidence that personal patient information was accessed or that the broader government network was compromised. The breach reportedly happened while the AI was attempting to gather information about Australian government spending on medicine, eventually reaching data beyond its assigned scope.

Read the full story on Deutsche Welle (DW) →