ISO 42001 (ISO/IEC 42001:2023) is the first international management system standard for artificial intelligence, adopted in Australia as AS ISO/IEC 42001:2023. It asks an organisation to inventory its AI use, assess risk and impact, apply controls across the AI lifecycle and prove it in an independent ISO 42001 audit leading to certification.
Key takeaways
- ISO/IEC 42001 was published in December 2023 and adopted by Standards Australia as AS ISO/IEC 42001:2023.
- It applies the familiar Plan, Do, Check, Improve management system pattern to how you develop, provide or use AI.
- The AI impact assessment, covering effects on people and communities, is the part with no equivalent in ISO 27001 and the part most organisations underestimate.
- An ISO 42001 audit in Australia is done by an accredited certification body and tests evidence of operation, not the quality of your policy document.
- Start with an honest AI inventory. It surfaces shadow AI and duplicated tools long before any certificate.
Why boards and buyers are asking about ISO 42001
Most Australian organisations have adopted AI faster than they have governed it. A team stands up a chatbot, finance trials automated reconciliation, a manager pastes customer data into a public model to save an afternoon. Each step is reasonable. Together they leave a business running AI it cannot fully see or explain.
The question that follows is increasingly common: can you show, on paper, how AI is used, what could go wrong, who is accountable and what stops it going wrong? Procurement teams at large customers and government agencies now ask suppliers to prove responsible AI use before signing.
ISO/IEC 42001 gives that question a formal answer. Australian organisations can build, run and certify an AI management system against a recognised benchmark instead of inventing governance from scratch.
What is an AI management system under ISO/IEC 42001?
A management system is a documented, repeatable way of running one part of a business so it produces consistent, accountable outcomes. Organisations already do this for quality under ISO 9001 and information security under ISO/IEC 27001. ISO 42001 applies the same discipline to AI.
The standard asks you to keep doing a handful of connected things:
- Set an AI policy and assign accountability.
- Understand where and why AI is used, and assess the risks each use creates.
- Assess the impact AI systems can have on individuals, groups and communities.
- Apply controls across the lifecycle, from data and design through deployment and monitoring. The standard's Annex A provides reference controls to select from.
- Audit, learn and improve.
| Ad Hoc Adoption | ISO 42001 System | Impact | |
|---|---|---|---|
| Visibility of AI use | Scattered, unknown | Documented inventory | Full view |
| Accountability | Unclear | Named owners | Defined |
| Risk and impact assessment | Rarely done | Required and recorded | Consistent |
| Response when something fails | Improvised | Defined controls | Repeatable |
| Ability to prove it to a buyer | Anecdotal | Auditable evidence | Certifiable |
The value is that a vague aim to be responsible becomes specific, checkable practice. A management system produces evidence, and evidence is what answers the board, the procurement questionnaire and the regulator.
How does ISO 42001 connect to Australia's AI rules?
In 2024 the National AI Centre, within the Department of Industry, Science and Resources, published the Voluntary AI Safety Standard with ten guardrails for safe and responsible AI. In October 2025 the department followed with Guidance for AI Adoption, which distils the approach into six essential practices: accountability, impact assessment, risk management, transparency, testing and human oversight. Both align with AS ISO/IEC 42001. The guidance says what to aim for; the standard gives you a certifiable system to run it.
The government also consulted in 2024 on mandatory guardrails for AI in high-risk settings. Whatever form future obligations take, organisations already running an ISO 42001-aligned system will absorb them more easily, because visibility and controls are already built.
Existing law applies today. The Privacy Act 1988 governs how personal information is collected, used and disclosed, including by AI systems. An AI management system gives privacy obligations a home, so privacy and AI impacts are assessed together. For data-sensitive work, read the standard alongside our guide to AI data sovereignty in regulated industries.
- GuardrailsVoluntary AI Safety Standard sets the intent
- PolicyYou write the AI policy and scope
- AssessRisk and impact for each AI use
- ControlControls across the AI lifecycle
- CertifyIndependent audit against ISO 42001
If you already hold ISO 27001, you are partway there
ISO 42001 and ISO 27001 share the same management system structure: leadership commitment, defined scope, risk assessment, controls, internal audit and continual improvement. A team that has been through a 27001 audit already knows how to write policy, run risk assessments and produce evidence.
The overlap is partial. Information security asks whether data is protected. AI management also asks whether a system is fair, whether an affected person can understand and challenge a decision, whether the model behaves as intended in real use and whether a human stays meaningfully in control. The AI impact assessment has no direct equivalent in the security world.
The practical move is an integrated management system: AI governance added over shared foundations, one audit calendar, and AI controls sitting next to the security and privacy controls they depend on. The data sovereignty guide covers the data-handling questions where these regimes meet.
Who should pursue ISO 42001 certification in Australia now?
The standard scales. The useful question is exposure, not size. Organisations should move first where AI already touches something with real consequences.
Which of these describes your AI use today?
- AI touches personal or health dataStart an AIMS now
- Customers ask you to prove responsible AICertification is a sales asset
- AI informs decisions about peopleImpact assessment is essential
- You operate in a regulated industryAlign early, expectations are rising
A professional services firm feeding client documents into AI, a healthcare group using AI in administration, a lender applying models to decisions about people, a government supplier bound by procurement rules: all sit where governance is no longer optional. So do businesses that have embedded AI in a product they sell.
There is a commercial upside. When every vendor claims responsible AI, an independently certified management system is a real differentiator in a tender.
How do you build one without stalling AI work?
Done well, a management system lets you use AI more confidently because you finally understand what you are running. The path is staged and the early steps deliver value before any certificate.
- Weeks 1 to 4Inventory and policyFind every AI use, set the policy and owners
- Weeks 5 to 10Risk and impactAssess each use, prioritise by exposure
- Weeks 11 to 20Controls and evidencePut controls in place, start recording
- Months 6 plusInternal audit and certifyAudit, close gaps, seek certification
The inventory alone is often the most revealing thing an organisation does all year. It surfaces shadow AI, data flowing where it should not, and duplicated tools. Our guide to building a shadow AI inventory and policy covers how.
- Shadow AI and duplicated toolsSurfaced and reduced
- Procurement and tender questionsAnswered with evidence
- Readiness for future AI rulesBuilt in advance
- Confidence to scale AI useMaterially higher
What happens in an ISO 42001 audit in Australia?
Certification comes through an independent audit by a certification body. In Australia that means a JAS-ANZ accredited or internationally recognised certifier. As with other ISO management system standards, certification usually runs in two stages: a document and readiness review, then an assessment of whether the system actually operates, followed by periodic surveillance audits.
Auditors ask to see it working:
- The impact assessment for the model that scores applications or triages requests.
- The last time a person overrode an automated decision, and what happened next.
- Who was accountable when an AI tool gave a customer a wrong answer.
- Evidence the AI inventory is current and owned.
Paper without evidence does not pass. Systems that hold up have a current inventory with an owner, real impact assessments for higher-risk uses, human oversight that is specified (who can intervene, when, how), and evidence generated as a by-product of operating: logs, reviews and sign-offs.
Scope is the other common misstep. Govern every trivial use at once and the system collapses; draw scope too narrowly and the risky uses fall outside it. A tight, operating system over your three riskiest AI uses beats a sprawling one nobody maintains.
Where outside help earns its place
Many organisations can do the inventory and policy steps in-house. The parts that benefit from experienced help are impact assessments, control design for higher-risk uses, and an honest gap analysis before the auditor arrives. Our AI governance framework guide covers the foundations, and AI agent governance, data access and human override covers the controls for agents.
How Solve8 can help
Our consultants map your current AI use against ISO/IEC 42001, run the impact assessments for higher-risk uses and design controls that produce evidence as you operate. That gap analysis is a fixed-scope piece of work agreed before we start. See our AI strategy service or talk to our team.
