Australia’s Voluntary AI Safety Standard and Guardrails

Australia’s National AI Centre published the Voluntary AI Safety Standard to give practical guidance for safe AI development and deployment. The publication sets out ten guardrails for organisations across the AI supply chain, including accountability and governance, risk management, data governance, testing and monitoring, human control, transparency, contestability, supplier/deployer transparency, incident management and stakeholder engagement. The department states that the guardrails are voluntary. The word “must” in an example or implementation prompt does not turn the publication itself into a statute. Contract terms, regulatory directions or other laws may separately create binding duties. The VAISS aligns its concepts with international frameworks such as ISO/IEC 42001 and NIST’s AI Risk Management Framework, but alignment is not certification and does not mean that following the document guarantees compliance with every law. Its intended audience includes developers, deployers and other organisations throughout the supply chain. Organisations can select measures proportionate to context, but should not use flexibility to avoid controls for higher-risk uses. On October 21, 2025, Australia released Guidance for AI Adoption, which condenses the ten guardrails into six essential practices and is described by the department as an updated and simplified evolution of the VAISS and AI Ethics Principles. The guidance separates organisation-wide governance from system-specific steps such as testing. This later publication does not make the VAISS mandatory or repeal it; it gives organisations a more accessible pathway. The 2025 National AI Plan continues to promote safe AI adoption through voluntary guidance and risk-based guardrails. As of September 26, 2026, distinguish these voluntary instruments from binding privacy, consumer, discrimination, safety and sector laws.