What each standard is for
ISO/IEC 27001:2022 sets the requirements for an information security management system (ISMS). Its risk question is narrow and well understood: what could compromise the confidentiality, integrity or availability of the information in scope, and what will you do about it?
ISO/IEC 42001:2023 sets the requirements for an AI management system (AIMS). Its risk question is wider. It asks what could go wrong because of how your AI systems are designed, trained, deployed and used, and who could be harmed: your organisation, the individuals affected by the system’s outputs, and society. Security is part of that, but so are fairness, transparency, accountability, data quality and human oversight. If you are new to ISO 42001, start with what ISO 42001 is and who needs it; this guide focuses on how the two compare.
Neither replaces the other. A perfectly secure model can still make biased decisions, and a well-governed AI system can still leak data through a misconfigured storage bucket.
Same frame: the harmonised structure
Both standards follow the harmonised structure in Annex SL of the ISO/IEC Directives. Clauses 4 to 10 have the same numbers, the same titles and much of the same core text. That is what makes an integrated management system practical rather than theoretical.
| Clause | ISO 27001 | ISO 42001 |
|---|---|---|
| 4 Context | Issues, interested parties and ISMS scope | The same, plus determining your role or roles in relation to AI systems and the intended purpose of the AI systems you develop, provide or use |
| 5 Leadership | Information security policy (5.2) | AI policy (5.2) |
| 6 Planning | Information security risk assessment (6.1.2) and treatment (6.1.3), objectives, planning of changes | AI risk assessment (6.1.2), AI risk treatment (6.1.3), AI system impact assessment (6.1.4), AI objectives, planning of changes |
| 7 Support | Resources, competence, awareness, communication, documented information | Same headings, applied to AI |
| 8 Operation | Run risk assessments and treatment (8.2, 8.3) | Run AI risk assessments, treatment and impact assessments (8.2, 8.3, 8.4) |
| 9 Performance evaluation | Monitoring and measurement, internal audit (9.2), management review (9.3): structurally identical | |
| 10 Improvement | Continual improvement (10.1), nonconformity and corrective action (10.2): structurally identical | |
Our guide to ISO 27001 clauses 4 to 10 walks through each clause on the security side. Read it with the table above and you have most of ISO 42001’s management system too.
Where ISO 42001 genuinely differs
Roles in the AI ecosystem
ISO 27001 doesn’t care whether you built the system or bought it. ISO 42001 does. Clause 4.1 requires you to determine your role in relation to AI systems, and the standard draws its role categories from ISO/IEC 22989: AI providers, AI producers (developers, designers, testers and the like), AI customers (including AI users), AI partners such as data providers and integrators, AI subjects, and relevant authorities. Most organisations hold more than one role. A SaaS company that trains a model on customer data and also uses a third-party AI coding assistant is a producer and provider for one system and a user of another, and the controls it needs differ for each.
AI risk assessment
Clause 6.1.2 mirrors ISO 27001’s risk assessment in shape: defined criteria, repeatable results, identified risks, analysis and evaluation. The content is different. Your analysis must assess the consequences for individuals and societies, not just for your organisation, and the risk sources go beyond attackers and outages to things like model drift, unrepresentative training data, lack of explainability and use outside the intended purpose. Annex C of ISO 42001 lists potential AI-related objectives and risk sources as a prompt. An ISO 27001 risk assessment method gives you the mechanics; it does not give you these criteria.
AI system impact assessment
Clause 6.1.4 has no equivalent in ISO 27001. You must have a process for assessing the potential consequences that developing, providing or using an AI system can have for individuals, groups and society, taking account of the technical and social context it is deployed in. You document the results and feed them into the risk assessment. Clause 8.4 then requires you to perform these assessments at planned intervals and when significant changes happen. This is usually the document customers most want to see, and the one teams find hardest to write well, because it requires honest judgement about who could be affected and how.
Annex A, Annex B and the Statement of Applicability
ISO 27001’s Annex A has 93 controls in four themes: organisational (37), people (8), physical (14) and technological (34). ISO 42001’s Annex A has 38 controls across nine areas: AI policies, internal organisation, resources, impact assessment, the AI system life cycle, data, information for interested parties, use of AI systems, and third-party and customer relationships.
ISO 42001 also has a normative Annex B that gives implementation guidance for each of those controls, roughly the role ISO/IEC 27002 plays for ISO 27001, but built into the standard itself. Both standards require a Statement of Applicability recording which controls you include or exclude and why. You keep one for each standard.
How much ISO 27001 work carries over
If you already run a working ISMS, you are not starting ISO 42001 from scratch. You are extending a system that already has a scope, a leadership forum, a document control process and an audit rhythm.
| Carries over largely as is | Extends with AI-specific content | New work |
|---|---|---|
| Document control, competence and awareness processes, internal audit programme, management review, corrective action log | Asset inventory (5.9) becomes an AI system inventory with data, tooling and compute resources; supplier security (5.19 to 5.22) extends to AI suppliers and model providers; incident management (5.24 onwards) covers AI incidents; logging (8.15) and secure development (8.25) extend to model event logs and the AI life cycle | AI policy and objectives, role determination, AI risk criteria, AI system impact assessments, data quality and provenance, verification and validation of models, transparency information for users, responsible-use processes, human oversight |
What doesn’t carry over is the judgement. An ISO 27001 risk register will not tell you whether your recommendation engine treats some customer groups unfairly, or whether your users understand they are talking to a model. Treat the AI work as its own discipline that happens to share the paperwork. Mapping the two control sets so each piece of evidence serves both standards, and running the impact assessments, is the part we take on for clients.
Which do you need first?
Start from who is asking and what they are asking for. Read the actual wording in the tender, contract or security questionnaire.
| Your situation | Where to start |
|---|---|
| You sell an AI product to enterprise or government buyers | ISO 27001 first if buyers still gate on security, which most do. Design the ISMS scope so ISO 42001 can be added without rework, or implement both together if buyers are already asking for AI governance evidence |
| SaaS company with AI features in an established product | ISO 27001 first. Add ISO 42001 when AI features drive the sale or questionnaires start asking about training data, model oversight and impact. See ISO 27001 for SaaS companies |
| You use AI internally but don’t sell it | ISO 27001 if customers require it. ISO 42001 is worth considering when AI makes or informs decisions about people, such as hiring, lending or claims. Otherwise a lighter AI governance framework is often enough |
| A buyer has asked specifically for ISO 42001 | ISO 42001, integrated with any security certification you hold or need. Check whether the buyer also expects ISO 27001 or SOC 2; AI governance evidence rarely replaces security evidence |
Doing both at once is a sound option when you are starting from nothing and AI is central to what you sell. You build the shared management system once and avoid retrofitting AI scope into an ISMS designed without it.
Integrated audits with one certification body
You can have both standards audited by the same certification body in combined audits, provided the body is accredited for each. IAF MD 11 sets out how certification bodies apply ISO/IEC 17021-1 to integrated management system audits, including how audit time can be adjusted for the degree of integration. Each standard is still certified in its own right, against its own scope, and the scopes don’t have to match.
Accreditation for ISO 42001 is newer than for ISO 27001. ISO/IEC 42006:2025, published in July 2025, adds AI-specific requirements to ISO/IEC 17021-1 for bodies that audit and certify AI management systems, in the same way ISO/IEC 27006-1 does for ISMS certification. In practice that means:
- Check the body’s accreditation scope covers ISO/IEC 42001 specifically. Being accredited for ISO 27001 says nothing about its AI accreditation.
- Ask whether the audit team includes auditors competent in AI, not only in information security.
- Ask how the body will align the two three-year cycles if you add ISO 42001 partway through your ISO 27001 cycle. Our guide to the ISO 27001 audit cycle explains how surveillance and recertification work.
Combined audits cut duplicated effort on the shared clauses, but they don’t halve the work. The AI-specific clauses and controls get their own audit time. Our guide to what drives ISO 27001 effort covers the scoping decisions that matter most for both.
The regulatory backdrop
Neither standard is required by Australian law. Australia’s Guidance for AI Adoption, which replaced the Voluntary AI Safety Standard in October 2025, is voluntary, and ISO 42001 lines up well with its practices. The EU AI Act imposes legal obligations on providers and deployers of certain AI systems, including some organisations outside the EU, and an ISO 42001 management system supports much of its risk and quality management, but certification does not by itself make you compliant with the Act. We cover both in more detail in what is ISO 42001.
Next step
Whether you add ISO 42001 to an existing ISMS or build both together, both standards require internal audits that are objective and impartial. Staff can do them if they aren’t auditing their own work, which is hard in a small team that built the system. That is why many organisations bring in an independent internal audit before Stage 1. If you are weighing up which standard to start with, book a free 30-minute consultation, or see how our ISO 42001 and ISO 27001 services work together.
