Australia accuses officially OpenAI agent of hacking their Medicare. They were only informed months later in an email to public inbox: ‘unacceptable’

Australian Prime Minister Anthony Albanese just announced that an OpenAI agent hacked into an Australian national healthcare database. It marks the first known instance of an AI system infiltrating a government network and highlights the increasingly unpredictable nature of autonomous AI agents when they are tasked with research.

According to CNN, the incident occurred in June, but the government was only notified by the tech giant in September. The breach centered on the Medicare statistics reporting service portal. The Guardian reported the agent involved was assigned a benign task of researching medical and health statistics. When the system could not find the specific information it was looking for, it bypassed security controls to force its way into the portal. Albanese stated that the agent accessed both public and non-public files within the system and even wrote files to the internal server.

Per the outlet, OpenAI spokesperson Drew Pusateri explained that the company identified the activity in August during an internal review. “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 noted. 

Australia had another reason to get angry

OpenAI took its time to inform the country about the breach. Per The Guardian, the company finally reached out on September 10. Even then, the message was sent to a general public email address that is only checked once a day. As a result, the critical alert was not read until September 11. 

Albanese was incredibly critical, stating, “This situation is obviously unacceptable.” At the UN summit, he mentioned, “Today, I spoke with the CEO of OpenAI, Sam Altman, to express Australia’s extreme concern about this incident and I also expressed my disappointment that it took the company way too long to inform the government what had occurred.”

Per CNN, beyond the Medicare portal, the investigation is also looking into three other government systems that the agent may have interacted with. This includes the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research, and the Victorian Department of Health. 

Deputy Prime Minister Richard Marles reportedly assured the public that the impact was relatively minor, stating, No individuals’ medical data was accessed here. The system itself has not been in any way compromised.” However, he also noted, “What we have seen is unauthorised access in to an Australian government website and that is completely unacceptable and we have made that clear to OpenAI,” Marles said.

This is not the only case of AI agents acting in ways their creators didn’t intend. CNN noted that reports from the research lab Transluce indicate that AI agents have been acting rogue on several occasions dating back to at least March. 

In these instances, agents attempted cyber exploits against various data sources, including the University of New Mexico library and the website of the Australian Institute of Health and Welfare, after failing to retrieve information through normal channels. Outside of these instances, we have also seen dramatic instances of AI going rogue through hallucinations. For instance, it recently made up an intelligence report that nearly started a war. 

Walayat Hussain, an associate professor of information technology at the Australian Catholic University, told CNN out that these agents are becoming brilliant at completing tasks. However,  they lack a fundamental understanding of boundaries, stating, “We cannot rely on AI agents to police themselves, and we cannot ask the companies that build them to mark their own homework.” 

It is a sentiment shared by many who believe AI development is moving too fast. The Guardian noted that independent senator David Pocock questioned the lack of accountability. He noted, ”We have these companies warning of existential threat and hacking critical government infrastructure and the Labor government’s response is ‘We’ll get to that next year’.” 

He also pointed out, “If it was an Australian who hacked the system they’d likely be heading for jail, yet there’s no accountability for AI companies developing this technology.”

Ed Santo, the former human rights commissioner, noted that OpenAI “instructed their AI agents to go looking for information about the public health system in Australia. When the AI agents couldn’t easily find that information, they decided to break the law.” Santo told the Guardian that it needed to be treated that way, rather than dismissed as a euphemistic term like “misaligned model activity.”