ISO/IEC 42001:2023 is the world’s first certifiable standard for an AI management system. It sets out how an organisation governs the AI it builds, supplies or uses: leadership, policies, an AI risk and impact assessment, human oversight and monitoring. Certifiable on a three year cycle, not legally mandatory in Australia.
What does ISO 42001 actually require?
A management system, not a one off test of a model. ISO/IEC 42001:2023, published in December 2023, asks an organisation to build, run and improve an AI management system, an AIMS, the same way ISO 27001 asks for an information security one. It runs on clauses 4 to 10, the Annex SL backbone it shares with ISO 27001 and ISO 9001: context, leadership, planning, support, operation, performance evaluation, improvement. Annex A adds 38 controls grouped under nine objectives, selected against your AI risk and impact assessment and recorded in a Statement of Applicability. Certification says nothing about how good any single model is. It certifies the governance around the AI, not the accuracy of the output.
The nine Annex A objectives, and what each governs:
| Annex A objective | What it governs |
|---|---|
| A.2 Policies related to AI | An AI policy set by leadership, aligned to the organisation’s objectives and risk appetite |
| A.3 Internal organisation | Roles, responsibilities and reporting lines for AI, including who owns each decision |
| A.4 Resources for AI systems | The data, tooling, compute and people an AI system depends on, documented and managed |
| A.5 Assessing impacts of AI systems | The AI system impact assessment: consequences for individuals, groups and society |
| A.6 AI system life cycle | Responsible design, development, testing, deployment and retirement of AI systems |
| A.7 Data for AI systems | How training and operational data is sourced, prepared and governed for quality and provenance |
| A.8 Information for interested parties | What you tell users, regulators and affected people about the AI and its limits |
| A.9 Use of AI systems | Responsible operation, including human oversight of automated decisions |
| A.10 Third party and customer relationships | Allocating responsibility across suppliers, model providers and customers |
Who needs ISO 42001?
Any organisation that develops, provides or uses AI, not only the labs that train models. A bank running a third party model for credit decisions, a SaaS company shipping an AI feature, a government team deploying a chatbot: each is in scope, because each carries responsibility for how the AI behaves once it reaches real people. The standard is deliberately broad on this. Demand is commercial and procurement driven. Customers, partners and tender panels have started asking how you govern AI, and a certificate is the cleanest way to answer. Why AI governance matters now sets out what is pushing that change.
What is the AI system impact assessment?
This is the requirement that separates ISO 42001 from a security standard. Alongside ordinary risk to the organisation, the standard makes you assess the consequences of an AI system for individuals, groups and society: fairness and bias, transparency, safety, accountability. Clause 6 requires it in planning; clause 8 requires you to perform and document it in operation. The shift is the point. A security control asks whether the system could be breached. The AI risk and impact assessment asks whether the system, working exactly as designed, could harm the people it touches. Those are different questions.
Is ISO 42001 mandatory in Australia?
No. There is no law that requires it. In September 2024 the Department of Industry, Science and Resources published a Voluntary AI Safety Standard with 10 guardrails covering accountability, risk management, data governance, testing, human oversight, transparency, contestability, supply chain transparency, record keeping and stakeholder engagement, and consulted on mandatory guardrails for AI in high risk settings. The guardrails line up closely with ISO 42001, so an AIMS is a practical route to meet them. The direction of mandatory regulation is still moving, so treat the voluntary standard as where the floor sits today, not where it stays. An ISO 42001 readiness checklist is a sensible first read.
How does ISO 42001 fit with ISO 27001 and the EU AI Act?
It bolts onto an existing management system. Because 42001 shares the Annex SL structure with ISO 27001, an organisation already certified to 27001 reuses much of the leadership, risk and improvement scaffolding and adds the AI controls and the impact assessment on top. It is not a security standard with AI attached, though. The EU AI Act is a different instrument again: binding regulation, in force from 1 August 2024 and phased through to 2027. ISO 42001 certification does not by itself make you compliant with the Act, and the Act does not require the certificate. The governance maps onto the Act’s risk and quality management expectations, so certification is a strong foundation, not a substitute for legal advice. ISO 42001 vs the EU AI Act compares the two in detail.
How long does it take, and what does it cost?
On Cybernion engagements, an AI governance gap assessment and AI system inventory runs about 3 to 6 weeks, and a full implementation 4 to 9 months, depending on how many AI use cases sit in scope and how mature your existing governance is. Those are indicative figures, not a quote, and the cost drivers matter more than any headline number. Certification, if you choose to pursue it, runs on the same three year cycle as ISO 27001: a Stage 1 and Stage 2 audit by an accredited body, then surveillance audits. Pricing is scoped by complexity and is not published. The larger cost is rarely the audit. It is building the AI inventory and the governance an organisation should have had before the auditor arrives. The complete ISO 42001 guide sets out the full picture, and ISO 42001 readiness is where Cybernion helps.
Frequently asked questions
Yes. Unlike an IRAP assessment, ISO 42001 is certifiable by an accredited certification body, on a three year cycle of a Stage 1 and Stage 2 audit followed by surveillance audits.
Not in Australia. Demand is commercial and procurement driven. The Voluntary AI Safety Standard of September 2024 is voluntary, though mandatory guardrails for high risk AI are under consultation.
ISO 27001 protects information. ISO 42001 governs AI, and adds an AI system impact assessment that looks at harm to individuals, groups and society, not only risk to the organisation.
No. The certificate is a strong governance foundation, but it is not legal compliance with the Act, and the Act does not require it.
Written by Gaurav Vikash, an ASD endorsed IRAP assessor and senior cyber security leader with 18 years of experience across Australia, the UK and Asia, including CISO and senior security leadership roles. He holds CISSP, CISA, CISM and CRISC and is an ISO 27001 and ISO 42001 Lead Implementer, and speaks regularly at industry conferences.
Sources:
- ISO/IEC 42001:2023, Information technology, Artificial intelligence, Management system, December 2023
- Voluntary AI Safety Standard, Department of Industry, Science and Resources, September 2024
- EU AI Act, European Commission, 2024
Last updated: 21 June, 2026