The report recommends that the Government:

Recommendation 1: The Office of AI should commission and publish a National AI Risk Assessment (NAIRA) 

The Office of AI should commission and publish a comprehensive assessment of Australia’s exposure to AI-related risk, considering the full range of AI risks — including prospective, cross-cutting and potentially catastrophic risks — and the effectiveness of Australia’s regulatory frameworks. This analysis would establish a shared basis for prioritising and funding reviews and reforms.

Recommendation 2: The Government should legislate mandatory monitoring and reporting of serious AI incidents

The Government should establish a mandatory monitoring and reporting mechanism for serious AI incidents, harms, and near misses, to give agencies the visibility needed to prioritise reforms. The National AI Plan commits the Australian AI Safety Institute (AISI) to monitor AI risks and harms, and regulators to identify, assess and address AI-related harms. AI incident monitoring and reporting would provide the evidence base to enable these bodies to do this work. Established models, in specific sectors in Australia and for AI overseas, could inform the design of such a mechanism. These include Australia’s reporting regimes for aviation, therapeutic goods, and ransomware payments, and the incident reporting obligations in the European Union’s AI Act and California’s Transparency in Frontier Artificial Intelligence Act.

Recommendation 3: The Department of Industry, Science and Resources (DISR) or the Office of AI should coordinate and publicly report on AI-related reviews and reforms

DISR or the Office of AI should be funded to coordinate and track the ongoing reviews and reforms of regulatory frameworks. This should also involve a public, whole-of-government register of AI-related reforms and reviews, and annual reports to Parliament.

Global Shield Australia has set out the benefits of such a scheme and how it would work. Read our brief here: AI Incident Monitoring & Reporting for Australia

Recommendation 4: The Joint Select Committee on AI should use its inquiry to examine options for mitigating risk created primarily at the model level of the AI supply chain

The Joint Select Committee on AI’s inquiry is a key opportunity to review the role of frontier AI developers and the regulatory interventions available to mitigate risk at the level of AI models, where controls can be most effective across multiple policy areas. The Committee should consider and make recommendations on Australia’s ability to regulate frontier AI developers, how best to address model-level risk, and options for Australia to contribute to international rules and norms on AI. The Committee should also review obligations legislated in overseas jurisdictions (including the European Union, South Korea, Vietnam, California and New York) and identify which may be suitable for Australia to adopt. This is particularly important to inform the design of the Australian Standards for AI, which are expected to come to Parliament in early 2027. 

Recommendation 5: The Government should provide specific funding for AI-related reviews and reforms and set clear timeframes for their conclusion

The Australian Government should fund priority review and reform processes, based on the NAIRA (Recommendation 1), with dedicated expertise and clear timeframes. Increased funding for the AISI would also enable it to better achieve its mandate to support agencies in their reviews and reforms.

Recommendation 6: The Government should enact framework or coordinating legislation to support the AI-related reviews and reforms

The Government should enact framework legislation setting baseline, cross-cutting expectations, definitions, and obligations to support the range of AI-related reviews and reforms taking place. This could include rules on attribution and liability for AI agents, information-sharing provisions, and foundational obligations (such as to manage risk when developing or deploying AI models). It could also include mandatory incident monitoring and reporting (Recommendation 2), obligations arising from the review of frontier AI developers (Recommendation 4), and empower a coordinating body to oversee and report on Australia’s AI-related reforms.

As the capabilities of AI systems continue to rapidly increase, risks that were once just speculative are becoming present dangers. This includes risks of catastrophic-level harms that cannot be easily responded to or recovered from. The National AI Plan’s third pillar commits the Government to “keep Australians safe”. This report provides an assessment of Australia’s progress towards meeting that aim and concrete actions to enable the Government to deliver on that commitment.