Briefing
Reference 3 · Primer · About 5 minutes
Download the MP3 · Leonard, an AI voice, reads a script written from the sources below.
Written from public sources and checked on 10 October 2026. The standard is paraphrased, not reproduced. Check the current text before relying on it.
The standards and frameworks that describe how to run AI governance as a system, and the practical work each one calls for.
These are Agile Insights AI Strategy & Governance offerings. Some are delivered with partners.
Sign up for our podcast at agile-insights.com.au.
Leonard This is the one international standard that lets an organisation prove, to an outside auditor, that it governs its artificial intelligence properly. It is called I S O, I E C, 42001, and by the end of this episode you will know what it asks of you, what a certificate means, and what to do first. This is our reference series, as at October 2026.
Leonard First, what it is. It is published by the International Organization for Standardization and the International Electrotechnical Commission, in December 2023. It is a management system standard for AI, often shortened to A I M S. It is voluntary, and it is certifiable. That means an accredited outside body can audit you against it and issue a certificate. It works the same way as I S O 27001 does for information security.
Leonard Now the problem it solves. Boards keep hearing "we have an AI policy". Customers and regulators want more than that. They want evidence that someone is accountable, that risks were assessed before a system went live, and that the controls still work a year later. Picture a procurement team asking a supplier how it manages AI risk. A one page policy is not an answer. A certificate is.
Leonard Who uses it? Chief executives and boards who need assurance. Risk, privacy and technology leaders who run the programme. And suppliers who sell into regulated buyers. It comes up in tenders, in vendor due diligence, in audit committee papers, and when a regulator asks how you govern AI. Certificates are becoming more common among suppliers.
Leonard What is in it? Four ideas carry it. One: it is a management system, so it follows a plan, do, check, act cycle. Leaders set the AI policy, define roles and own the result. Two: it is risk based. You assess your AI risks, then choose controls and record, in a document called the Statement of Applicability, which ones you apply and why. Three: it requires an impact assessment. You must consider what an AI system could do to people and society, not only to your business. Four: it demands proof of operation. Internal audits, management reviews and corrective action are required, and certified organisations are re-audited every year. Many organisations do the early steps well and skip the checking. That is a binder, not a management system.
Leonard The standard also carries a list of reference controls, called Annex A. It is usually counted as about 38 controls in nine groups. They cover policy, accountability, resources, impact assessment, the system life cycle, data, information for users, how systems are used, and suppliers and customers. Think of it as a menu, not a checklist.
Leonard How does it connect to the rest of the stack? It sits in the governance layer. Your data foundations sit beneath it, because the standard expects data quality and provenance. Local law sits above it. It does not replace the EU AI Act or Australian privacy law, and it does not test whether a particular model is safe or fair. It was published before agentic systems became common, so agent specific controls need to be added on top.
Leonard A few close relatives are worth knowing. I S O, I E C, 22989 is the shared vocabulary. 23894 is guidance on AI risk management. 42005 is guidance on running the impact assessments, published in May 2025. And 42006, published in July 2025, sets the rules for the bodies that audit and certify against 42001. All of these are published.
Leonard Getting started takes three steps. First, build a register of every AI system you use or build, with a named owner for each. Second, run a gap analysis against the standard, to see what you already have from existing security or quality certifications. Third, draft the impact assessment process and the Statement of Applicability, then test them on one real system. To learn more, start with the standard itself, the Standards Australia pages, and the guidance on impact assessment. The show notes link to them.
Leonard Now currency, as at October 2026. The 2023 edition is still the current text. As far as the public catalogue shows, no revision or amendment has been registered. Australia adopted it unchanged as an Australian Standard in February 2024. On certification, major bodies are now accredited. B S I announced accreditation from the American accreditation board in March 2026, and Intertek has held accreditation from the Australian and New Zealand accreditation body since July 2025. A separate draft, 42007, on certification schemes for AI systems, is still in public enquiry. Always check the current text, and your certifier's accreditation, before relying on it.
Leonard So, in short: 42001 turns AI governance into something an outsider can audit. If you'd like to know more, sign up for our podcast at agile-insights.com.au.
All of the reference series (16) · Follow AI Governance Reference in Apple Podcasts or any podcast app
Everything from Agile Insights: briefings, Making You Think and the reference series