InsightsAI governance
ISO 42001 explained: what an AI management system certification involves and who needs it

Key takeaways
- ISO/IEC 42001 is the first certifiable standard for managing AI responsibly.
- It builds on the ISO 27001 structure and adds an AI policy, AI risk and impact assessments, and 38 AI-specific controls.
- Start with an inventory of every AI system you build, sell or use.
A new question has started showing up in security questionnaires: "How do you govern your use of AI?" For companies building AI into their products, or using it in decisions about customers and employees, there is now a certifiable answer. ISO/IEC 42001 is the international standard for managing AI responsibly, and it is quickly becoming the AI equivalent of ISO 27001.
What ISO 42001 is
ISO/IEC 42001 was published in December 2023 by ISO and the IEC. It sets requirements for an AI management system: the policies, roles, risk processes and controls an organization uses to develop, provide or use AI responsibly. Like ISO 27001, it is a management system standard, which means an accredited auditor checks that the system exists, that it runs, and that it improves over time.
It uses the same structure as ISO 27001, with requirements in clauses 4 through 10. On top of that, Annex A lists 38 AI-specific controls across nine areas, covering AI policy, roles, resources, impact assessment, the AI lifecycle, data, information for users, how AI is used, and third parties.
What is different from ISO 27001
If you already run ISO 27001 or SOC 2, much of the structure will feel familiar. The AI-specific parts are where the work is:
- An AI policy that states how the organization will develop and use AI, approved by leadership.
- An AI risk assessment that looks beyond security to fairness, transparency, reliability and misuse.
- An AI system impact assessment that considers how each AI system could affect individuals, groups and society before it is deployed.
- Lifecycle controls for how AI systems are designed, tested, deployed, monitored and retired.
- Data controls for where training and input data come from, its quality and how it is handled.
- Human oversight and transparency, so people know when AI is involved and someone is accountable for its outputs.
- Supplier controls, because much of the AI a company uses arrives inside other vendors' products.
Who needs it
- Companies selling AI products to enterprises, especially in healthcare, finance and other regulated markets, where buyers increasingly ask for proof of AI governance.
- Companies using AI in decisions about people, such as hiring, lending, care or pricing, where the stakes and the legal exposure are highest.
- Companies preparing for the EU AI Act. ISO 42001 lines up well with the Act's management expectations, though certification alone does not make you legally compliant.
- Companies that already hold ISO 27001 and want to extend a proven management system to AI rather than start a separate program.
If your use of AI is limited to staff using approved chatbots and writing assistants, you probably do not need certification yet. A clear AI use policy and AI checks in your vendor reviews will cover you for now.
How certification works
- Inventory your AI. List every AI system you build, sell or use, including AI features inside existing software.
- Set the scope. Decide which products, systems and teams are in. A tight scope keeps the first certification manageable.
- Run a gap assessment against the clauses and Annex A controls.
- Build the core documents: AI policy, risk and impact assessment methods, and a Statement of Applicability that explains which Annex A controls apply and why.
- Run the system for a period: complete assessments, train staff, monitor AI systems and record decisions.
- Do an internal audit and a management review.
- Complete the certification audit, in two stages, with an accredited certification body. Certificates last three years, with surveillance audits in between.
Companies that already have ISO 27001 can often reuse their document control, internal audit and management review processes, which shortens the path considerably.
Common mistakes
- No inventory. You cannot govern AI you have not found, and AI features arrive quietly through software updates.
- Treating it as paperwork. Auditors look for evidence that impact assessments and monitoring actually happen.
- Skipping the impact assessment or folding it into the security risk assessment. It asks different questions.
- Ignoring vendors. If a supplier's AI makes or shapes decisions for you, its risks are yours too.
How Sollucio can help
Sollucio's AI governance service takes companies from an AI inventory and gap assessment through to certification readiness, and connects ISO 42001 to the ISO 27001, SOC 2 and HIPAA work you may already have, so you build one program instead of several.
Next step
Is a customer asking how you govern AI?
We will tell you whether ISO 42001 makes sense for you now, and what a realistic path looks like.


