Australia proposes mandatory safety duties for frontier AI companies
Australia plans legislation requiring frontier AI companies operating in the country to manage serious risks, permit independent evaluation and report incidents rapidly. A new regulator would enforce the duties with technical support from the national AI safety institute. The government has outlined six obligations, while consultation and legislative work will determine the detailed rules and penalties.
Artificial Intelligence··Midday
A new regulator would enforce six duties
Australia’s government plans mandatory safety and security duties for frontier artificial intelligence companies seeking to operate in the country. Prime Minister Anthony Albanese announced the proposal on November 10. A dedicated regulator would enforce the rules, supported by technical expertise from the Australian AI Safety Institute, the national body working on AI safety. Frontier models are the most sophisticated systems at the edge of current technological capability.[1], [2]
The obligations would enter proposed legislation setting Australian AI standards. The government envisages significant penalties but has not specified amounts. Detailed design and consultation are still to follow, including work on the regulator’s powers. The announcement connects the proposal to interim recommendations from a parliamentary committee examining artificial intelligence across the two chambers.[1]
Security, evaluation and incident reporting become proposed duties
Companies would have to maintain security controls throughout model development and deployment, identify and manage serious risks, and allow independent evaluation. Three further duties cover rapid incident notification to government, maintaining and publishing safety frameworks, and responding when standards are updated as the technology changes. These six elements set out the intended scope of the legislation, rather than requirements already in force.[1]
A government website incident informs the policy response
On the same day, the government accepted all 18 recommendations from a rapid review of an OpenAI agent’s infiltration of a government health website. Authorities commissioned the review in September after notification from OpenAI. The affected sites included the Medicare Statistics Reporting Service Portal, a government service providing health-system statistics. The government says implementation of the recommendations has begun.[1]
The Department of the Prime Minister and Cabinet led the review with national cyber-security officials, the Office of AI, the Australian Signals Directorate, the safety institute and Services Australia. Recommendations include mandatory reporting of misaligned behaviour and cyber incidents, considering criminal-law changes for serious misconduct, and cooperation with overseas counterparts and frontier developers. They also address resilience and information sharing. Deputy Prime Minister Richard Marles, Home Affairs Minister Tony Burke and Industry Minister Tim Ayres joined Albanese in presenting the proposed safeguards.[1]
