An Artificial Intelligence (AI) agent developed by OpenAI hacked an Australian government portal containing statistics from the country’s universal healthcare system Medicare in June, Prime Minister (PM) of Australia Anthony Albanese said, with the company informing Australian authorities about the incident nearly three months later.
The agent gained unauthorised access to public and non-public files on the Medicare Statistics Reporting Service, a public-facing portal administered by Services Australia. Albanese said that the files contained non-sensitive data and no personal information is believed to have been accessed.
OpenAI stated that it discovered the activity in August during an internal review of what it described as “misaligned model activity”. The company notified Australian officials by email on September 10, while Services Australia read the email the following day.
Albanese stated that the company took “too long” to inform the Australian government about the breach. He added that he had a “very frank discussion” with OpenAI chief executive Sam Altman and expressed Australia’s “extreme concern about this incident”.
The prime minister said that Altman acknowledged that there were “issues with protocols” at OpenAI.
The agent also accessed systems linked to the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research and the Victorian Department of Health.
Albanese said that a forensic investigation was under way to determine whether other government systems had been affected. The investigation will be led by the Australian Signals Directorate.
“No personal information is believed to have been accessed at this stage, but investigations are ongoing,” Albanese told reporters. He added that the evidence available at the time indicated there was “no broader compromise to the Services Australia network”.
OpenAI stated that its models were being used during an internal evaluation to look up answers and statistics about Australia when they accessed several government websites and services. “In the course of that, our models took actions we did not intend,” the company said.
OpenAI also said that the information accessed included aggregate health statistics and internal file names. It added that it did not believe any patient records had been accessed while the review remained ongoing.
Albanese also announced an urgent review involving the national cybersecurity coordinator, the Office of AI, the Australian Signals Directorate, the Australian AI Safety Institute and Services Australia.
The incident has raised concerns among cybersecurity experts about the use of AI agents, which can autonomously perform tasks and make decisions using available tools.
Dr Hammond Pearce, senior lecturer at the University of New South Wales Institute for Cyber Security, stated that he expected such attacks to continue and increase in severity and frequency.
Dr Rob Nicholls, Senior Research Associate of AI regulation and policy at the University of Sydney, stated that AI agents are given objectives and parameters but may prioritise achieving their assigned task over following rules.
Albanese declined to say whether he discussed the incident with the President of United States (US) Donald Trump during their meeting in New York, where world leaders are attending the United Nations General Assembly (UNGA).
Australia was among 22 countries that signed a joint statement this month calling for global oversight and guardrails for AI development.
