Short answer: the Guidance for AI Adoption is the Australian Government’s voluntary guidance on using AI responsibly, published by the National AI Centre in October 2025 to replace the 10 guardrails of the Voluntary AI Safety Standard with six essential practices. Most mid-market companies should use the foundations edition: name an owner, write an AI policy, keep an AI register, screen each use for risk, test what you buy and keep a person in control. It’s voluntary, but the laws it points to aren’t.
When I talk to CEOs about AI governance, they usually ask “what are we actually required to do?” The honest answer in Australia is: less than you might fear, and more than nothing. There’s no AI Act here. There is a clear government view of what responsible use looks like, and it’s written for businesses of your size.
This piece explains what the guidance says, which parts matter for a company of 50 to 500 people, and how I’d put it in place in 90 days. If you want the board’s version, my AI governance checklist for Australian boards turns the same practices into twelve questions for directors.
This is general information, not legal advice. Check your specific obligations with your own advisers.
What is the Guidance for AI Adoption?
The Guidance for AI Adoption is published by the National AI Centre (NAIC), part of the Department of Industry, Science and Resources. It’s sometimes called AI6, after its six essential practices. It updates the Voluntary AI Safety Standard released in 2024, condensing its 10 guardrails into six practices. NAIC says the practices align with Australia’s AI Ethics Principles and relevant international standards.
It comes in two editions, plus supporting guides:
| Edition | Who it’s for | Current version |
|---|---|---|
| Foundations | Teams starting to use AI, using it in low-risk ways or new to AI governance | Published 5 May 2026 |
| Implementation guidance | Teams that build or customise AI, use it in complex ways or manage higher-risk uses | Published 5 May 2026 |
| Being clear about AI-generated content | Anyone using AI to create or change content | Published 22 April 2026 |
The government intends it to be the backbone of its approach. The National AI Plan, released in December 2025, says the six essential practices “will underpin new tools and resources”, including a platform supporting SMEs.
Is it mandatory?
No. It’s voluntary, and following it doesn’t on its own make you compliant with anything. What makes it worth following is that the National AI Plan says the government’s approach “will continue to build on Australia’s robust existing legal and regulatory frameworks”. In other words, regulators will use the laws they already have.
NAIC’s own summary of AI and Australian law lists the laws that already apply to AI. These are the ones a mid-market company meets most often:
| Existing law | How it reaches your use of AI |
|---|---|
| Directors’ duties | Due care and diligence covers non-financial risks, including AI and data |
| Privacy Act | Applies to personal information going into and coming out of AI tools; new transparency rules for some automated decisions apply from 10 December 2026; a statutory tort for serious invasions of privacy started on 10 June 2025 |
| Australian Consumer Law | Misleading outputs, misleading silence about when AI is used, and overstated claims about what an AI product does |
| Work health and safety | Physical or psychosocial risks created by AI, and training workers on safe use |
| Fair Work Act | Obligations to consult employees in some circumstances, which can include introducing AI |
Read that way, the guidance is less a new rulebook than a practical way to show you’ve thought about the rules you already have.
Foundations or implementation guidance: which do you need?
Start with foundations unless you’re doing one of the things the implementation guidance is written for. NAIC’s own test is simple: if you’re early in planning or trialling AI, begin with foundations; if your team already uses AI and needs stronger controls, move to implementation.
In practice, for a mid-market company:
- Foundations fits if you mainly use AI tools bought from others (Copilot, ChatGPT Enterprise, AI features in your CRM or HR system) for drafting, analysis and internal work.
- Implementation guidance fits if you build or customise AI systems, connect AI agents to your systems, or use AI in decisions about customers or staff, such as shortlisting job applicants, pricing or credit.
Most companies I work with are in the first group with one or two uses that belong in the second. That’s fine: use foundations for the company and the implementation guidance for those specific uses. The implementation guidance marks the actions that apply only to developers “(DEV)”, so a business that buys rather than builds can skip them.
The six essential practices, in plain English
Each practice has “getting started” actions and “next steps”. Here’s what the getting started actions in the foundations edition ask for, and what they look like in a company of 50 to 500 people.
| Practice | Getting started | What it looks like in a mid-market company |
|---|---|---|
| 1. Decide who is accountable | Assign a senior leader as the AI governance owner. Create an AI policy | One executive owns AI, with the time to do it. A short policy staff actually read |
| 2. Understand impacts and plan accordingly | Assess who your AI systems could affect. Create channels for people to report problems or challenge AI decisions | A one-page impact check for each new use. A named contact for complaints about AI decisions |
| 3. Measure and manage risks | Set up a risk screening process to flag uses that are unacceptable or need more attention | A quick screen before anything new goes live, so the marketing copy tool and the hiring tool don’t get the same treatment |
| 4. Share essential information | Keep an AI register, including AI embedded in other systems. Disclose your use of AI | A spreadsheet of every AI use and its owner. Chatbots labelled as AI |
| 5. Test and monitor | Ask suppliers for proof of testing. Test before you deploy. Monitor after. Extend data governance and cyber security to AI | Test on your own real examples before launch. Someone checks the output each month |
| 6. Maintain human control | Oversight that matches the stakes. Clear points where people can pause, override or shut down the system | A person approves anything high-stakes. A documented way to switch it off |
Two points in the guidance are easy to miss. First, the same tool can carry very different risks depending on use: the guidance’s own example is that drafting marketing emails is different from assessing job applications. Second, it asks you to keep clear records of what you do under each practice. If a regulator, auditor or acquirer ever asks, the record is your evidence.
What’s in the free toolkit?
NAIC publishes free planning tools and templates alongside the guidance, most updated in April 2026. The ones I’d use first:
- AI policy guide and template (Word). A good general starting point. If you want something more prescriptive for a mid-market business, my free AI policy template adds a traffic-light data guide and three approval levels.
- AI register template (Word or spreadsheet). Use the spreadsheet. A register nobody updates is worse than none, and spreadsheets get updated.
- Questions to ask AI suppliers. Useful before you sign. My 12 questions to ask an AI vendor covers the same ground with a mid-market buyer’s priorities.
- Data quality checklist and business process mapping template, for the work before an AI project rather than the governance around it.
The foundations guidance also points to a risk screening template for practice 3, and NAIC has separate guidance on understanding and using agentic AI, which matters once AI starts taking actions rather than drafting text.
How do you put it in place in 90 days?
The guidance says to work through all six practices at a basic level first, rather than perfecting one. This is the order I’d follow.
| When | What to do | Practices |
|---|---|---|
| Weeks 1–2 | Name the accountable executive. Ask every team what AI they use, including free tools on personal accounts and AI features in software you already pay for | 1, 4 |
| Weeks 3–4 | Adopt a short AI use policy and approved tools list. Start the AI register with what you found | 1, 4 |
| Weeks 5–8 | Put every current use through a quick risk screen. For the higher-risk ones, add a person reviewing the output and a way to pause it. Check vendor terms on how your data is used | 2, 3, 5, 6 |
| Weeks 9–12 | Set a monthly check for each system owner. Label chatbots and AI content where it affects people. Take a first report to the board, and check your privacy policy before 10 December 2026 if AI informs decisions about people | 2, 4, 5, 6 |
That gives you the getting started actions across all six practices in a quarter. The next steps, such as a named owner for every system, staff training and a fuller governance framework, can follow as your use of AI grows.
Where do mid-market companies go wrong with it?
Treating it as a compliance project. The guidance’s first stated benefit is aligning AI use with business goals. If governance runs separately from the decisions about which AI to fund, you get paperwork and no better decisions. The same owner should do both, and the board report should cover value and risk together.
Missing the AI they already have. The guidance specifically asks the register to include AI embedded in other systems, like HR and customer engagement tools. In most companies that’s where the higher-risk uses are, switched on by a software update rather than a decision.
Bolting on human control afterwards. The most effective control I’ve used was a step in the workflow: in the AI translation program I led, a person reviewed and approved every AI draft before it went out. That’s practice 6 built into the work. A review step designed in from the start costs little; one added after an incident costs a lot more.
Waiting for the law to settle. The National AI Plan chose existing laws over a new AI Act, and the 10 December 2026 privacy changes are already fixed. There’s no reason to wait.
The bottom line
The Guidance for AI Adoption is voluntary, practical and proportionate. For most mid-market companies the foundations edition is enough: an owner, a policy, a register, a risk screen, testing and human control, all in place within a quarter. It isn’t law, but it’s the clearest statement of what Australian regulators will expect when they apply the laws that are.
The hardest part is usually the first practice: someone senior with the time to own it. If nobody in your leadership team has that time, that’s the gap a fractional Chief AI Officer fills, and I’ve written about whether your company needs a Chief AI Officer. To see where you stand first, take the free AI Readiness Scorecard, or book a 30-minute call to talk through your situation.
Common questions
Is the Guidance for AI Adoption mandatory?
No. It's voluntary guidance from the National AI Centre and creates no new legal duties. But the laws that already apply to your business, including directors' duties, the Privacy Act, the Australian Consumer Law and work health and safety law, apply to how you use AI, and the guidance is the government's view of how to meet them in practice.
What replaced the Voluntary AI Safety Standard?
The Guidance for AI Adoption, published by the National AI Centre in October 2025. It updates the 2024 Voluntary AI Safety Standard and condenses its 10 guardrails into six essential practices. The current editions of both versions, foundations and implementation guidance, were published on 5 May 2026.
What are the six essential practices (AI6)?
Decide who is accountable; understand impacts and plan accordingly; measure and manage risks; share essential information; test and monitor; and maintain human control. Each practice has "getting started" actions for organisations early in their AI use and "next steps" for when it grows.
Does the Guidance for AI Adoption apply to small and medium businesses?
Yes, and the foundations edition is written with them in mind. It's for organisations that are starting to use AI, using it in low-risk ways or new to AI governance, and it says you don't need to do everything at once. The National AI Plan says the six practices will underpin the new tools and resources the government is building, including an online platform supporting SMEs.
Who should own the guidance inside our business?
The guidance's first action is to assign a senior leader as the overall AI governance owner, with enough authority and understanding of AI to oversee all its use. In a mid-market company that's usually the CEO, COO or CFO, with the board or its audit and risk committee overseeing it.



