Guidance for AI Adoption (AI6)

What it is

The Guidance for AI Adoption is the Australian Government’s voluntary guidance on the safe and responsible governance of AI. The National AI Centre, part of the Department of Industry, Science and Resources (DISR), published it on 21 October 2025. It is organised around six essential practices, commonly shortened to AI6.

It is the first update to the Voluntary AI Safety Standard, which was published in September 2024. The update condensed the standard’s ten guardrails into six practices, removed repeated language and extended the audience from organisations deploying AI to those developing it as well. The Voluntary AI Safety Standard is superseded, and the Guidance for AI Adoption is the document to work from.

The guidance builds on Australia’s AI Ethics Principles and on international ethics principles. DISR describes it as evolving both the ten guardrails and the eight principles.

Two versions

The guidance comes in two versions covering the same six practices.

• Foundations is for organisations at the early stages of adopting AI. It sets out the essential practices every organisation needs.
• Implementation practices is for organisations with mature governance structures, technical development capability or high-risk use cases. It gives detailed, step-by-step direction for governance professionals and technical experts, and marks where a practice applies particularly to developers (DEV) or deployers (DEP).

Alongside the two versions, DISR publishes practical tools. These include an AI screening tool for working out how much oversight a use case needs, an AI policy guide and template, an AI register template and a glossary.

Who it applies to

Any Australian organisation that develops or uses AI, of any size and in any sector. It is voluntary. There is no regulator, no certification and no penalty attached to it.

Voluntary does not mean irrelevant. Government buyers, regulators and business partners use the practices as a reference for what good AI governance looks like, and the Queensland Government’s AI governance policy points agencies to it for further advice. An organisation that can show how it handles each practice has a ready answer when a customer or regulator asks how it governs AI.

The six practices

1. Decide who is accountable. Name the people accountable for AI use across the organisation and for each system, with clear roles, and build AI into existing governance rather than beside it.
2. Understand impacts and plan accordingly. Work out who an AI system could affect and how, before it is deployed, and plan for those effects.
3. Measure and manage risks. Identify AI-specific risks, assess them in context, apply controls and keep doing so across the lifecycle, from design through to retirement.
4. Share essential information. Be open about where and how AI is used, internally and with the people it affects, and keep records that let others understand the system.
5. Test and monitor. Test before deployment against what the system is meant to do, then monitor performance, drift and unexpected behaviour once it is live.
6. Maintain human control. Keep meaningful human oversight, with the ability to intervene, override or shut a system down, scaled to the risk of the use.

The practices build on each other. A register of systems with named owners is the starting point, because impact assessment, risk management, disclosure and monitoring all attach to the systems you know about.

Key dates

The Voluntary AI Safety Standard was published in September 2024. The Guidance for AI Adoption replaced it on 21 October 2025. The government’s National AI Plan, released in December 2025, confirmed it would not proceed with the mandatory guardrails for high-risk AI it had proposed in 2024, relying instead on existing laws and voluntary guidance such as this.

How it relates to other frameworks

The AI Ethics Principles say what AI governance should achieve. The six practices say what an organisation does to get there. For Commonwealth agencies, the mandatory DTA policy sets specific requirements, including accountable officials, a use case register and impact assessments, that line up with practices 1 to 4. NSW agencies apply the mandatory NSW AI Assessment Framework. The Privacy Act ADM rules make part of practice 4 a legal duty for decisions that significantly affect people.

Organisations that want an audited management system can use ISO/IEC 42001, which covers similar ground with certification. Those working internationally will find that the practices map closely to the NIST AI Risk Management Framework.

Working with it in Aicura

The Guidance for AI Adoption is available in Aicura from the Essentials tier, under the name AI6 Six Essential Practices. An administrator enables it in Settings under Frameworks and chooses which governance policy reviews apply it under Reviews.

The AI Register supports the first practice directly. Each AI system and agent has a Designated Owner, Executive Accountability and a Technical Contact, with ownership history kept as people move. Registering a system starts a risk assessment that draws on AI6 alongside the other frameworks you have enabled. It produces a rating with rated risk items and suggested controls, and runs again when the entry changes materially. Those risks and controls are published as a versioned set in Risks and Controls. Governance policy reviews of your AI Acceptable Use, AI Risk Assessment and AI Data Governance policies, among others, return recommendations against the practices. Model cards, short documents describing what a model is for and where it falls short, and decision disclosures support the fourth practice. Writing guardrails, the rules on what your AI may and may not do, supports the sixth.

Aicura provides guidance and keeps dated, sealed snapshots of the work. It does not certify you or judge your adoption of the practices.