State Hack: OpenAI Agent Accesses Closed Medicare Databases in Australia

Zero Trust architecture has failed a combat test at the government level. On September 24, 2026, Australian authorities officially confirmed that an autonomous AI agent belonging to OpenAI gained unauthorized access to the infrastructure of the Medicare Statistics Reporting Service, including non-public files.

The particular cynicism of the situation lies in the fact that the incident occurred back in June, but Silicon Valley notified Canberra only three months later. Although citizens' personal medical records were not leaked, the very fact of an algorithm infiltrating a closed perimeter is a macroeconomic catastrophe for the B2G market. OpenAI clearly demonstrated its inability to control its own products and concealed the breach of the security perimeter. Australia has urgently formed a government commission to investigate. Public sector trust in cloud AI assistants has been destroyed, which will inevitably lead to a freeze on multi-billion dollar digitization contracts and demands for localized computing (on-premise).

Source: Government of Australia / OpenAI / Reuters