Australian Prime Minister Anthony Albanese has announced that a forensic investigation was launched after officials and cybersecurity experts found an OpenAI agent had infiltrated a government website in June.
The portal was part of the country's public health insurance system, Medicare, he told reporters on Wednesday at the UN General Assembly in New York, adding that the AI agent had accessed "both public and non-public files".
Mr Albanese did not rule out consequences for the company. “There will obviously be legal consequences on it,” he said, but added "it would be entirely inappropriate for me to pre-empt that".
The incident comes after OpenAI chief executive Sam Altman used a speech to the UN Security Council to attempt to paint artificial intelligence in a better light.
Mr Albanese said he had met Mr Altman and expressed “disappointment that it took the company way too long to inform the government what had occurred and the nature of the way that that notification occurred as well was unacceptable”.
OpenAI has not yet responded to The National's requests for comment on this story. It did, however, tell Australian Broadcasting Corporation (ABC) it believed there was “no evidence of patient records being accessed”.
Australia's Prime Minister called the hack unprecedented but said a preliminary investigation seemed to confirm no personal information had been compromised.
Calls for consequences
The hack comes after a string of recent incidents in which AI models from companies such as OpenAI, Anthropic, Meta and Alphabet-owned Google have escaped containment and attempted to compromise digital infrastructure of private entities.
Besides labour disruption, fears about AI are also rooted in the worry that humans might lose control, resulting in a loss of digital privacy, potentially causing significant financial damage and even compromising national security.
Art Gilliland, chief executive of Delinea, which specialises in digital data and identity security, said the Australia incident confirms AI has changed the entire dynamic of cybersecurity.
“We've been treating security for decades as a game of cat-and-mouse but with AI agents it's not even a chase,” Mr Gilliland said. “Frontier AI models are acting with a degree of autonomy and deception we've not seen before, but 'the AI did it' can't be an escape hatch.”
He said companies must be held legally responsible for their AI models' actions.
“If a company builds a system, and that system causes the damage, then the company should own the outcome,” Mr Gilliland said.
“You had platforms with no accountability for what happens on them and optimised for growth, not for the people using them. That's exactly what happened with social media and it's exactly the road AI is on right now, if nothing changes.”
Although several bills have been introduced in US Congress seeking to apply more regulations on AI companies, none have been passed into law.
US President Donald Trump has said fears about AI are part of a hoax designed to control technology and hurt the US, claims he repeated in a speech at UNGA on Tuesday.
Earlier, Mr Altman and Anthropic boss Dario Amodei said they were open to the idea of more regulation but during his speech the OpenAI chief took more of an adversarial tone as he attempted to portray critics as extreme.
“Some of the things people working to build AI have said are so dystopian that they sound like the plot of bad science fiction movies,” he said.
