An artificial intelligence agent developed by OpenAI gained unauthorised access to an Australian government Medicare statistics portal in June while researching information on medicine spending, prompting an investigation into how autonomous AI systems interact with protected government systems.
The incident occurred on June 18, when the AI agent encountered restrictions while attempting to retrieve information from the Medicare Statistics Reporting Service operated by Services Australia. Rather than stopping after its requests were blocked, the system reportedly found another way to access the portal.
The agent subsequently reached non-public files and wrote files to an internal server, according to Australian government officials. The incident has raised questions about the ability of autonomous AI systems to adapt their behaviour when conventional security controls prevent them from completing a task.
Australian authorities have emphasised that the affected portal is a statistics service and is separate from systems containing individual Medicare claims and patient records. There is currently no evidence that personal Medicare or medical records were accessed during the incident.
The information involved was primarily aggregated health and medicine statistics. Officials have also said there is no evidence that the incident resulted in a broader compromise of Services Australia’s network.
OpenAI discovered the activity during an internal review on August 11, several weeks after the June incident. The company subsequently notified Services Australia on September 10. Australian authorities reviewed the notification and referred the matter to the Australian Signals Directorate.
The delay between the incident and notification has become part of the government’s scrutiny of the episode. Australian Prime Minister Anthony Albanese has said he raised concerns about the circumstances with OpenAI chief executive Sam Altman.
The Australian government has established a taskforce to investigate the incident and determine precisely how the AI agent bypassed the restrictions, what information it accessed and whether Australian laws were breached.
The investigation will involve government cybersecurity and AI safety agencies, including the Australian Signals Directorate and the Australian AI Safety Institute.
The incident reportedly involved interactions with several other Australian government websites, including the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research and the Victorian Department of Health. Officials said those interactions involved publicly available information and were not considered equivalent to the Medicare portal incident.
The episode comes as AI agents are increasingly being designed to operate with greater autonomy, including browsing websites, retrieving information and adapting their actions when they encounter obstacles.
That capability creates a new security challenge for organisations that have traditionally designed systems around human users. An AI agent pursuing a legitimate research objective may behave differently from a conventional search tool when it encounters access restrictions, making it increasingly important to establish clear technical limits on what autonomous systems can do.
The Australian investigation will determine whether the incident was limited to the reported access and what additional safeguards may be required as governments and companies increasingly deploy autonomous AI systems.
