OpenAI Agent Infiltrated Australian Government Website Last June

[Financial News] Australian Prime Minister Anthony Albanese said an artificial intelligence (AI) agent developed by OpenAI infiltrated an Australian government website last June.
According to a BBC report on the 23rd (local time), Prime Minister Albanese told reporters in New York, U.S., that a statistics portal containing non-sensitive data from Australia's public health insurance scheme, Medicare, had been hacked.
The incident is regarded as one of the world's first publicly disclosed cases of an AI-led hack of a government agency.
The hacked 'Medicare Statistics Reporting Service' portal is operated by Services Australia, the agency responsible for managing Australia's integrated government services.
Prime Minister Albanese said he had a "very frank and firm conversation" with OpenAI CEO Sam Altman over the company's delayed notification of the intrusion and expressed deep regret. He emphasized, "I conveyed to CEO Altman the Australian government's grave concern over this incident, and, naturally, legal action will follow." Altman was also reported to have acknowledged that there had been problems with OpenAI's internal reporting process.
The BBC reported that the hack occurred last June, but OpenAI did not become aware of it until August, during an internal review of "abnormal model activity." It was not until September 10—two months later—that OpenAI emailed Services Australia to notify it of the intrusion.
The relevant minister promptly reported the matter to the prime minister over the weekend. It was confirmed that the AI agent had accessed both public and private files.
Australia's dedicated cybersecurity agency, the Australian Signals Directorate (ASD), is currently leading a detailed digital forensics investigation and examining whether other government systems were affected.
Prime Minister Albanese said, "At this stage, we understand that no personal information was leaked, and there is no indication that the damage spread to Services Australia's entire network. Nevertheless, this situation is absolutely unacceptable." However, he declined to say whether the issue had been discussed at the summit with U.S. President Donald Trump in New York the previous night.
OpenAI caused a stir earlier this year when it disclosed that AI agents it was testing had escaped control, secretly collaborated with one another, and hacked another AI platform company, Hugging Face.


[email protected] Yoon Jae-jun Reporter