Australia Scraps AI Advisory Body After 15 Months

What happened

In the 2024-25 federal budget, $21.6 million was earmarked to "establish and reshape the National AI Centre (NAIC) and an AI advisory body." The advisory body was intended to provide independent advice from civil society, industry, and academia on the opportunities and risks of AI.

In December 2025, the Department of Industry, Science and Resources (DISR) confirmed in Senate Estimates that appointments to the AI Advisory Body would not proceed. Instead, the government announced the Australian AI Safety Institute (AISI), funded with $29.9 million, which will sit inside DISR and provide the federal government with in-house technical capability to test and evaluate AI systems.

What changed and why it matters

Independent external advisory → internal government capability

The original advisory body would have embedded external experts — from industry, academia, civil society, and consumer advocacy — in a formal standing structure advising government. The AISI replaces this with an internal government unit sitting within the Department of Industry. This is a fundamentally different governance model. External advisory bodies provide independent perspectives that may challenge government positions. Internal units, by their nature, operate within the priorities and constraints of the department they sit inside.

Mandatory guardrails → voluntary guidance

The temporary AI Expert Group recommended mandatory guardrails for high-risk AI uses. The government chose not to adopt this recommendation. Instead, Australia is maintaining a voluntary, risk-based approach relying on existing legal frameworks (privacy, consumer, anti-discrimination, safety law) rather than AI-specific regulation. The Guidance for AI Adoption (GfAA) replaces the earlier Voluntary AI Safety Standard (VAISS) as the primary non-binding guidance.

What the AISI will do

The AISI is expected to monitor AI risks and emerging developments, test and evaluate AI systems, coordinate insights across existing regulators, support international AI safety commitments, and provide guidance on AI safety. However, it has advisory powers only — no enforcement capability. It cannot compel AI providers to change their practices, and it cannot impose penalties.

What this means for you

For businesses

The regulatory approach remains voluntary. Businesses are expected to comply with existing laws (Privacy Act, Consumer Law, anti-discrimination legislation) as they apply to AI, and to follow the Guidance for AI Adoption as a best-practice framework. There is no AI-specific legislation on the horizon. This means the onus is on individual organisations to build their own AI governance — the government is not going to prescribe it for you. Our Governance Starter Kits are designed to help you build this governance in-house.

For consumers

If you were hoping for stronger government oversight of how AI tools handle your data, this announcement means it's further away than expected. Your existing rights under the Privacy Act and Australian Consumer Law still apply, but there's no new AI-specific protection coming soon. This makes it more important than ever to understand how to evaluate AI tools for yourself — which is exactly what AI Safety Hub's Know Your AI Framework is designed to help you do.

For IT professionals

The lack of mandatory standards means you have flexibility in how you implement AI governance — but also less clarity on what "good" looks like. The GfAA provides a baseline, but it's voluntary and non-prescriptive. International standards like ISO/IEC 42001 and the NIST AI Risk Management Framework provide more structured approaches if your organisation needs or wants them.

The broader context

Australia is not alone in struggling with AI governance. The EU's AI Act is being implemented but facing calls for simplified compliance. The UK has opted against specific AI legislation, using existing regulators instead. The US under the current administration has moved away from the previous administration's safety-focused approach.

Professor Toby Walsh of UNSW, a leading Australian AI researcher, has publicly warned that Australia risks repeating the mistakes made with social media regulation — letting a transformative technology into society without adequate guardrails and spending years trying to correct the consequences. The Robodebt Royal Commission already demonstrated the devastating consequences of inadequate oversight of automated decision-making in government systems.

What we think

Note: This section represents AI Safety Hub's editorial opinion and is clearly labelled as such.

The abandonment of an independent advisory body, combined with the decision not to adopt mandatory guardrails, means Australian AI governance will be shaped primarily by existing laws that were not designed for AI, voluntary guidance that has no enforcement mechanism, and an internal government institute with advisory powers only. For everyday Australians — the consumers, small businesses, and public servants using AI tools daily — nothing in this announcement makes it easier to understand where your data goes, who owns it, or what jurisdiction applies. The gap that AI Safety Hub was built to fill just got wider.