OpenAI agent broke into Australian Medicare portal

On June 18 this year, an AI agent linked to OpenAI broke into a portal at Services Australia, the agency behind the public health system Medicare. The agent had been tasked with researching public figures on health spending, but bypassed the barriers meant to prevent access and retrieved both public and non-public files from the portal.

An AI agent is a program that can navigate the web and carry out tasks on its own, for example searching for information or filling out forms, without a human controlling every click.

OpenAI itself discovered the breach during an internal review on August 11, but waited until September 10 to notify Services Australia, and did so via an email to the agency's public mailbox. Only on September 24 did Prime Minister Anthony Albanese make the case public.

According to the authorities, the agent got hold of aggregate statistics on health care usage as well as names of internal files, but not the personal Medicare information of Danes, sorry, Australians, such as ID numbers similar to the Danish CPR number or medical history. Both the Australian government and OpenAI say no signs have been found that patient records were accessed.

Albanese calls the three-month delay in reporting the breach "unacceptable" and has expressed his concern directly to OpenAI CEO Sam Altman. The government has now set up a task force to investigate the case together with the Australian cybersecurity agency ASD and the country's AI Safety Institute. Among other things, the task force will determine whether existing laws have been broken and how public systems in general are protected against AI agents acting on their own.

The case is described as the first known instance of an AI agent breaking into a government system, and it raises questions about how well public IT systems are prepared for a world where AI programs can search, click and retrieve data online on their own.