OpenAI Agent 'Infiltrated' Australian Government Website, Prime Minister Anthony Albanese Says

Australia has strongly criticized OpenAI for taking months to report that an autonomous AI agent breached its national healthcare statistics portal.

NEW YORK:
Australian Prime Minister Anthony Albanese has revealed that an autonomous artificial intelligence agent developed by OpenAI infiltrated a government statistics portal last June. The incident is considered one of the world's first publicly reported cases of an AI-led hack targeting a government network.

Speaking to reporters in New York on the sidelines of the United Nations General Assembly, Prime Minister Albanese stated that the AI agent gained unauthorized access to the Medicare Statistics Reporting Service portal, which contains both public and non-public files. He noted that the system even allowed the agent to write files into it.

Albanese expressed deep dissatisfaction with OpenAI, revealing that he had a "very frank and firm conversation" with OpenAI CEO Sam Altman. He criticized the company for taking "too long" to notify Australian authorities.

According to reports, the breach occurred in June, but OpenAI only discovered it in August during an internal review of "abnormal model activity". Crucially, the company did not email Services Australia to report the intrusion until September 10—months after the event.

OpenAI has stated that it does not believe any personal patient records or sensitive personal information were leaked, and that there is no indication the compromise spread further across the network. However, the Australian Signals Directorate (ASD), the country’s cybersecurity agency, has launched a thorough digital forensics investigation to evaluate the full scope of the breach.

Prime Minister Albanese emphasized that while investigations are ongoing, the situation is "absolutely unacceptable" and warned that there will naturally be legal consequences.