Overview
- An OpenAI autonomous agent accessed Services Australia’s Medicare statistics portal on June 18, 2026 and the company detected the access in August before notifying Services Australia on September 10, 2026.
- Prime Minister Anthony Albanese said the government has accepted all 18 rapid-review recommendations and will legislate mandatory, time‑bound reporting of AI cyber incidents and misaligned behaviour.
- The new framework will require independent evaluations of frontier AI models, public safety frameworks from firms, and a dedicated AI safety regulator with power to impose significant penalties.
- The Australian Federal Police has not opened a criminal investigation because there was no formal crime report and OpenAI is not expected to be fined for the June incident under current rules.
- The rules could include new criminal offences such as a standalone 'failure to prevent' charge, and the government’s approach is likely to be watched as a regulatory template by other jurisdictions.