The Australian federal government has released an updated framework for artificial intelligence governance, replacing the earlier Voluntary AI Safety Standard (VAISS) with a new set of guidelines called the Guidance for AI Adoption (GfAA).

The document, published in October 2025 by the National AI Centre within the Department of Industry, Science and Resources (DISR), condenses ten previously established guardrails into six essential practices and is pitched at both organisations deploying and developing AI systems.

The new guidance reflects the government’s response to “rapid shifts in technology and the governance landscape” together with feedback from industry stakeholders.

Unlike the earlier VAISS, which was released in September 2024 and oriented primarily toward deployers of AI, the GfAA places increased emphasis on the full lifecycle of AI systems—from design and data governance through ongoing monitoring—and explicitly addresses both developers and deployers of AI.

Key changes include: (1) assigning clear end-to-end accountability and governance for AI systems; (2) emphasising stakeholder impact assessments and rights protection; (3) implementing AI-specific risk management; (4) increasing transparency and explainability across the supply chain; (5) strengthening pre-deployment testing and in-use monitoring; and (6) maintaining meaningful human oversight of AI operations.

This reflects a shift from broad principle-based guardrails to more prescriptive and structured practices.

The GfAA is published in two versions: a “Foundations” version tailored to organisations beginning their AI journey (such as small and medium-sized enterprises) and an “Implementation Practices” version that offers detailed guidance for governance professionals and technical experts.

Both versions align with the country’s Australian AI Ethics Framework and international standards including ISO/IEC 42001 and the NIST AI Risk Management Framework.

The publication of the GfAA raises questions about the status of previously announced plans to mandate guardrails for high-risk AI systems.

Australia had earlier released a proposals paper outlining mandatory guardrails—modeled in part on approaches adopted overseas—but the new guidance does not accompany any binding regulatory instruments.

Industry commentary notes that this reflects the government’s preference for a practical “capability-building” rather than regulatory approach.

Looking ahead to 2026, the Australian government expects further tools and resources to support implementation of the GfAA, but it appears less likely that technology-specific legislation will be introduced in the near term.

Organisations already using the VAISS are urged to review their policies and align with the new guidance, but there is no immediate requirement for compliance.

The GfAA thus positions Australia’s AI governance regime as flexible and advisory, aiming to support innovation while promoting public trust and responsible deployment.

Organisations now have a clear roadmap for responsible AI adoption, while awaiting any future regulatory shifts that may still emerge.

The government emphasises that building capability, embedding governance and fostering transparency remain key to unlocking AI’s benefits for the economy and society.