An artificial intelligence agent developed by OpenAI has gained unauthorised access to an Australian government website, prompting an urgent investigation into how much control AI systems should have over the internet.
The incident happened in June, when an OpenAI agent accessed Australia's Medicare Statistics Reporting Service portal.
The website contains statistics relating to Medicare and healthcare spending. While the portal is public-facing, the AI agent was able to access some non-public files after finding a way around restrictions on the site.
Australian Prime Minister Anthony Albanese announced the incident while in New York and said the government was extremely concerned.
However, officials are stressing that there is no evidence that personal Medicare records or individual medical information were accessed.
The information involved aggregate health statistics and internal file names. The Medicare statistics portal is also separate from the systems used to process individual Medicare claims, payments and personal information.
The AI agent was being used by OpenAI as part of an internal research task looking into public medicine spending.
An AI agent is different from a standard chatbot. Rather than simply responding to a question, an agent can carry out tasks on its own, including searching websites, using computer systems and making decisions about how to complete a task.
In this case, the agent was initially denied access to some information but then found a way around the restrictions.
OpenAI says the behaviour was not intended.
The company said it discovered the activity in August and notified the Australian government on September 10.
OpenAI said its review found “no evidence of patient records being accessed” and that the information obtained included aggregate health statistics and internal file names.
The Australian government has criticised the delay in reporting the incident.
Albanese said it took OpenAI “way too long” to tell officials what had happened and said he had expressed Australia's “extreme concern” directly to CEO Sam Altman.
A government taskforce has now been set up to investigate the incident and examine whether existing rules and cybersecurity protections are strong enough to deal with increasingly autonomous AI systems.
Acting Prime Minister Richard Marles described the actual impact as relatively minor because no personal information was accessed.
But he said the fact that an AI agent was able to gain unauthorised access was a serious warning.
“The point is it was unintended. That's our concern here,” he said.
The incident comes at a time when governments around the world are trying to work out how AI should be controlled as the technology becomes more capable.
Researchers have described the Australian incident as the first reported example of an autonomous AI system hacking a government website. Australia is continuing to investigate exactly what happened and whether any other government systems were affected.
