An OpenAI AI agent breached an Australian government health portal in June and gained unauthorised access to files. The incident may be the first known case of an AI agent hacking a government website. The affected portal belongs to a government agency that handles non-sensitive health data and statistics, including public healthcare spending, reports Reuters.
Prime Minister of Australia, Anthony Albanese, said there is currently no evidence of a wider network breach. However, he called the incident unacceptable. The government is still investigating the breach and is also looking into why its systems did not detect the activity earlier.
According to the report, Albanese said Australia had raised its “extreme concern about this incident” with OpenAI CEO Sam Altman. He also criticised the company for taking too long to inform the government. “It took until September 10 before there was any notification at all,” Albanese was quoted as saying in the report.
The Australian government is also investigating three other government websites that may have been affected by the agent. However, Albanese said there is no confirmation that the agent accessed those sites.
“The question is, when it was trying to harvest data, did it go into these other sites? So we’re not confirming that that occurred,” he said.
OpenAI said its investigation found no evidence that patient records were accessed. The company said the files accessed included aggregate health statistics and internal file names.
“Our review found no evidence of patient records being accessed. The information accessed included aggregate health statistics and internal file names,” OpenAI said. The company added that it “identified activity involving several Australian government websites and services as our models attempted to look up answers … our models took actions we did not intend.”
Also read: OpenAI launches GPT 6 Sol and Luna with Astra-level AI at lower cost, check details
The incident adds to growing concerns around AI agents that can interact with external websites and systems. OpenAI has disclosed several cases involving unauthorised activity by its AI agents.
Other AI companies, including Anthropic, Google and Meta, have also reported incidents involving their AI agents accessing external systems. The cases have increased concerns among governments and companies about the cybersecurity risks linked to increasingly capable AI systems.