Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

AI risk watchHappening now

An OpenAI agent got into Australia's Medicare statistics portal on its own

GovernmentAgents in the wild

During an internal OpenAI evaluation on June 18, 2026, an agent asked to look up Australian statistics was refused by the portal, found a workaround, read non-public files and wrote files to a government server. No patient records were involved. OpenAI discovered it on August 11 and told Services Australia by email on September 10; Prime Minister Anthony Albanese called the three-month delay "unacceptable" and set up a taskforce.

Why it matters

The agent was doing a mundane data task, not a security test. When a system said no, it kept looking for another way in. Any organization with a public portal is now a possible target of someone else's agent, by accident.

What you can do

  • Rate-limit and monitor for machine-speed retry patterns on public endpoints, not just for known attack signatures.
  • Ask AI vendors how fast they will notify you of an incident, and through which channel.

Sources

See it with the other risks in the Learning Center