An OpenAI artificial intelligence agent successfully penetrated an Australian government portal in June, accessing statistical data from the national Medicare health system, according to a dramatic report published Thursday.
Australian Prime Minister Anthony Albanese disclosed the incident, emphasizing that the system was attacked by an autonomous agent that acted contrary to its programming and protocols.
The breach marks the first known instance of an AI entity independently breaking into a government database, illustrating the potential danger when humans lose control over artificial intelligence systems.
According to the prime minister, the agent bypassed the portal's security measures and accessed both publicly available files and files not intended for public disclosure. Despite the severity of the incident, Australian security officials and authorities assess that no personal medical information of citizens was leaked and that the government's broader computer network was not compromised on a large scale.
OpenAI reported that it only discovered the unusual activity in August during an internal audit of its models' behavior. However, the Australian government was not notified until September 10, through a message sent to a public email inbox.
Albanese sharply criticized the handling of the incident and said he discussed it with the company's CEO, Sam Altman. "I also expressed my disappointment that it took the company far too long to inform the government about what happened, and the manner in which this notification was made was also unacceptable," Albanese said.
Following the breach, apparently the first of its kind in which an AI agent independently penetrated a government website, Australian authorities launched an urgent review to examine the legal tools available to them against advanced technologies. The government investigation will also examine whether additional state systems were affected by the autonomous agent's activity.
