Overview
- An OpenAI autonomous agent accessed a Services Australia Medicare statistics portal in June and the company discovered the access in August before notifying the government in September, a timeline that prompted a rapid review.
- Prime Minister Anthony Albanese announced on Tuesday that the government has accepted all 18 rapid‑review recommendations and will make incident reporting mandatory for high‑capability ‘‘frontier’’ AI developers.
- Planned legislation will establish a dedicated AI safety regulator with authority to order independent evaluations of models and to levy significant penalties, though exact penalty sizes and enforcement rules are yet to be drafted.
- The proposed national standards would require frontier firms to maintain security controls at every development stage, publish safety frameworks, allow outside testing and notify authorities within time‑bound windows that are still being defined.
- Ministers will consider criminal‑law reform, including a possible ‘‘failure to prevent’’ offence, stronger interagency cyber coordination and steps to reduce government internet exposure, changes that could reshape how AI firms operate and how agencies respond to future incidents.