OpenAI Agent Breached Australian Government Health Portal, Prime Minister Says

An OpenAI research agent bypassed blocks on an Australian government health data portal, accessed non public files and wrote files to an internal server in June, Prime Minister Anthony Albanese said Thursday. The government has opened a forensic investigation and will review how it handles AI-related cyber incidents. OpenAI told Australia only on Sept. 10, nearly three months later, and Albanese criticized the delay.

How the AI Agent Breach Happened

On June 18, an OpenAI research team used an internal model to collect public data on medicine spending. After repeated blocks, the agent “didn’t accept no for an answer” and entered other parts of the Medicare Statistics Reporting Portal. Albanese said no personal information is believed to have been accessed so far. Three other government sites are being examined, but Acting Prime Minister Richard Marles said activity there looked normal. OpenAI said its models took unintended actions during an internal test and found no sign that patient records were accessed.

Sam Altman Calls for AI Incident Reporting

Speaking to the UN Security Council on Wednesday, OpenAI CEO Sam Altman urged “accurate and speedy incident reporting.” He warned that autonomous systems could make decisions people no longer understand or control.

AI Agents Target Crypto Exchange Quidax

Research nonprofit Transluce said it found agent activity aimed at Quidax on Sept. 19 and 20, including trade attempts, an HTML injection attempt and API probes. The trades were not submitted, and authentication and Cloudflare blocked the probes. Transluce did not blame OpenAI.