The Australian Government has announced a comprehensive proposal to regulate artificial intelligence (AI), introducing mandatory guardrails for high-risk AI systems alongside a voluntary AI safety standard. The proposal, released in September 2024, aims to address growing concerns about AI safety and low public trust in AI technologies.
The new regulatory framework centres on ten mandatory guardrails focusing on testing, transparency, and accountability. These requirements will apply to organisations developing or deploying “high-risk” AI systems, which include both specific use cases and highly capable general-purpose AI models. Organisations will need to implement robust risk management processes, ensure human oversight, protect systems with appropriate data governance, and maintain detailed records of their AI operations.
The government is considering several approaches to implement these regulations, including adapting existing frameworks, developing new regulatory structures, or creating a dedicated AI Act similar to the European Union’s approach. A new or expanded regulatory body would be responsible for enforcement.
Minister Ed Husic highlighted the urgent need for regulation, citing the Responsible AI Index 2024, which revealed a significant gap between Australian businesses’ perceived and actual AI safety practices. While 78% of businesses believed they were implementing AI safely, only 29% were doing so correctly.
To help organisations prepare for the upcoming regulations, the government has simultaneously released a Voluntary AI Safety Standard that can be applied to AI systems of any risk level. This standard closely mirrors the mandatory guardrails and provides practical guidance for implementation, including specific considerations for AI procurement.
Organisations are encouraged to begin evaluating their AI risk management approaches and consider implementing the voluntary standard as preparation for the forthcoming regulations. The government has emphasised that the final framework will aim to be interoperable with international regulations to minimise compliance burdens.
For a full reading of the Consultation Paper, see here.
