Australia: OpenAI agent infiltrated a government website

Australia is extremely concerned and disappointed by the security breach of a government website by OpenAI, which the artificial intelligence firm took too long to disclose, Prime Minister Anthony Albanese said. Albanese went public with the breach following a telephone conversation with OpenAI’s chief executive, Sam Altman.

Both are currently in New York to address the United Nations. “I spoke to Altman today to express Australia’s deep concern regarding this incident,” Albanese told reporters. “I also expressed my disappointment that the company took too long to inform the government of what had happened and that the manner in which the notification was made was also unacceptable,” Albanese added. An OpenAI agent had gained unauthorised access to the Australian Department of Health’s website on 18 July, he said.

OpenAI stated in a statement that, during a review of activities involving several Australian government departments, the company discovered that “our models had taken actions that were not our intention”. OpenAI notified Australia of the breach via an email sent on 10 September to a generic government department email address, Mr Albanese explained. On Thursday, Australia announced the launch of an investigation into the security breach.

The investigation will assess whether OpenAI could potentially face prosecution, Mr Albanese said. The investigation will also look into why Australian security agencies failed to detect the breach before OpenAI disclosed it. It is believed that no personal information was accessed, Albanese said. The AI agent had gained unauthorised access to a public portal dedicated to Medicare statistics. Albanese said he believed there were commercial reasons behind the AI’s investigation into how much was being spent on certain medicines and where spending patterns were changing.

Deputy Prime Minister Richard Marles stated that the incident represents the first known case in which an artificial intelligence agent has gained unauthorised access to the Australian Government’s IT systems. “The impact, if you can call it that, of this incident is relatively minor, as no personal information was accessed. The system itself was not compromised,” Marles told the Australian Broadcasting Corp. “That said, this is a very serious incident. What we are dealing with is an artificial intelligence agent that, unintentionally, gained unauthorised access to an Australian government website,” he added.