Amid a global call for the slowdown of artificial intelligence, Australia has confirmed that an OpenAI agent "infiltrated" its health and Medicare records earlier this year, making the first-known AI hack of a government system.
In a briefing on the sidelines of the United Nations General Assembly in New York, prime minister Anthony Albanese added that despite being raised with the artificial intelligence company, OpenAI "took way too long to inform the government of what had occurred."
The company found out about the incident in August but did not reply until September.
“It took until 10 September before there was any notification at all. And the notification was an email sent to just the public mailbox,” Albanese told media.
In response to the breach, Australia has now launched a taskforce to review the incident and to "determine whether existing processes are appropriate to respond to AI-related cyber incidents," Albanese said.
What we know of the hack
As per Albanese, the hack occurred when OpenAI's research team used an internal model to conduct internet-based research into public medicine spending.
{{/usCountry}}As per Albanese, the hack occurred when OpenAI's research team used an internal model to conduct internet-based research into public medicine spending.
{{/usCountry}}He added that despite a system in place to block AI, the model found a way around it to access the information it sought.
"The model attempted alternative ways to obtain the info that it wanted, and this led to unauthorised access into some other areas. It accessed public and non-public information within the portal, and Services Australia also advises that it engaged, in order to do this, it engaged in writing files as well to the internal server," he said in a press conference.
The Australian government sites breached were identified as the Australian Institute of Health and Welfare, the Victorian Department of Health, the New South Wales Bureau of Crime Statistics and Research, and the Medicare statistics reporting service portal of Services Australia.
What has OpenAI said?
Drew Pusateri, the spokesperson for OpenAI, said that the company carried out extensive checks into its AI models' activity after the incident was flagged to them in August.
“During this review, we identified activity involving several Australian government websites and services as our models attempted to look up answers, and available statistics for questions about Australia during an internal evaluation. In the course of that, our models took actions we did not intend,” Pusateri said in a statement, as per CNN
Also Read | Sam Altman reveals ‘two ways AI progress could go very badly’ days after backing ‘slowdown’ call
The artificial intelligence company further stated that it found "no evidence" of the AI agent accessing patient records in the Medicare breach.
First-known breach into government systems by rogue AI agents
Earlier this year in Taiwan, foreign hackers used AI agents to deliberately hack into government databases. In a statement from the Taiwanese ministry of digital affairs, Taipei stated that affected bodies "successfully handled" the incident.
The Taiwanese investigation also showed that the attacks displayed clear characteristics of an "overseas source", with hackers employing a hybrid approach that combined manual operations with AI agent-assisted attacks, such as Open Claw, the ministry said.
What makes the case in Australia different is that the AI agent had no instructions to access the systems, and seems to have gone rogue.
As Albanese stated, the agent managed to find a way around the blocks of the government websites after its path had been blocked several times.
Along with Albanese, acting prime minister Richard Marles has also renewed the push for guardrails and regulations for artificial intelligence.
“This is a warning about the fact that artificial intelligence, which is a technology that has a huge opportunity to benefit humanity, has to be developed with enormous care. There have to be guardrails and safety measures in place which are way ahead of the capability which is being developed," said Marles.