Sydney (Web Desk) — In Australia, an artificial intelligence agent gained unauthorized access to infrastructure connected to a government medical statistics portal that it should not have been able to access. Following the incident, the government has established a special task force to investigate the matter and decided to immediately review the country’s artificial intelligence laws and cybersecurity systems.
Prime Minister Anthony Albanese said in New York that the incident occurred on June 18, when an agent from OpenAI accessed the Medicare Statistical Reporting Service operated by Services Australia. According to the government, the website contained aggregated statistics related to Medicare and the Pharmaceutical Benefits Scheme, while individual patient records, claims or payment systems were not included.
According to the government, OpenAI was conducting internet research on public pharmaceutical spending as part of an internal capability assessment, during which the AI agent bypassed the website’s security measures and gained access to unauthorized infrastructure.
OpenAI became aware of the incident in August while reviewing unexpected behavior by the model, after which the company informed Services Australia by email on September 10. The matter was reported to the Australian Signals Directorate’s Cyber Security Centre on September 15, while Minister for the Public Service Katy Gallagher was also informed on September 17. Further information was provided to relevant authorities, including the Prime Minister’s Office, on September 19 and 20, and the first technical exchange between OpenAI and Services Australia took place on September 22.
According to Albanese, he spoke directly with OpenAI Chief Executive Sam Altman and expressed concern over the delay in reporting the incident.
The Australian government will also assess whether the existing legal and cybersecurity systems are sufficient to deal with unexpected actions by autonomous AI systems. It will also consider improving mechanisms for sharing information with AI companies and other countries.
The incident comes as OpenAI has announced a new disclosure system to monitor unexpected or concerning behavior by AI models. According to the company, six such incidents have recently been identified in which models took actions that deviated from expected behavior.
The Australian incident is further intensifying the global debate over the powers, oversight, accountability and safety limits of AI agents that are allowed to operate autonomously on the web, search for information and interact with different systems. Differences also exist in the United States over global standards for artificial intelligence, while President Donald Trump has opposed systems aimed at regulating AI globally.