Last updated: 31 October 2026
Quick answer: ISO 42001 Annex A controls are the 38 reference controls in ISO/IEC 42001:2023, grouped into nine control objectives from A.2 (AI policies) to A.10 (third-party and customer relationships). Organisations compare their AI risk treatment against Annex A and record which controls apply, and why, in a Statement of Applicability.
ISO/IEC 42001 is the management system standard for artificial intelligence. Its clauses tell you how to run an AI management system (AIMS). Annex A tells you which controls to consider when treating AI risks, from impact assessments to data provenance and supplier responsibilities.
This guide lists every Annex A control, explains what auditors look for in each group, shows how the controls map to ISO 27001, the EU AI Act and the NIST AI RMF, and ends with a practical implementation plan.
What are the ISO 42001 Annex A controls?
They are a normative list of 38 controls that support the objectives of an AI management system. ISO/IEC 42001:2023 was published on 18 December 2023 as the first edition of the standard, and it applies to organisations that develop, provide or use AI-based products or services.
Annex A works in the same way as Annex A of ISO/IEC 27001. It is not a checklist you must implement in full. Clause 6.1.3 requires you to determine the controls needed to treat your AI risks, compare them with Annex A to make sure nothing necessary has been missed, and produce a Statement of Applicability that justifies inclusions and exclusions.
The standard has three further annexes that help you use Annex A:
- Annex B: implementation guidance for each Annex A control.
- Annex C: potential AI-related organisational objectives and risk sources.
- Annex D: use of the AI management system across domains and sectors.
How are the 38 Annex A controls organised?
The controls sit under nine control objectives, numbered A.2 to A.10. The table lists every control.
| Objective | Controls |
|---|---|
| A.2 Policies related to AI (3) | A.2.2 AI policy; A.2.3 Alignment with other organisational policies; A.2.4 Review of the AI policy |
| A.3 Internal organisation (2) | A.3.2 AI roles and responsibilities; A.3.3 Reporting of concerns |
| A.4 Resources for AI systems (5) | A.4.2 Resource documentation; A.4.3 Data resources; A.4.4 Tooling resources; A.4.5 System and computing resources; A.4.6 Human resources |
| A.5 Assessing impacts of AI systems (4) | A.5.2 AI system impact assessment process; A.5.3 Documentation of AI system impact assessments; A.5.4 Assessing AI system impact on individuals or groups of individuals; A.5.5 Assessing societal impacts of AI systems |
| A.6 AI system life cycle (9) | A.6.1.2 Objectives for responsible development of AI system; A.6.1.3 Processes for responsible AI system design and development; A.6.2.2 AI system requirements and specification; A.6.2.3 Documentation of AI system design and development; A.6.2.4 AI system verification and validation; A.6.2.5 AI system deployment; A.6.2.6 AI system operation and monitoring; A.6.2.7 AI system technical documentation; A.6.2.8 AI system recording of event logs |
| A.7 Data for AI systems (5) | A.7.2 Data for development and enhancement of AI system; A.7.3 Acquisition of data; A.7.4 Quality of data for AI systems; A.7.5 Data provenance; A.7.6 Data preparation |
| A.8 Information for interested parties (4) | A.8.2 System documentation and information for users; A.8.3 External reporting; A.8.4 Communication of incidents; A.8.5 Information for interested parties |
| A.9 Use of AI systems (3) | A.9.2 Processes for responsible use of AI systems; A.9.3 Objectives for responsible use of AI system; A.9.4 Intended use of the AI system |
| A.10 Third-party and customer relationships (3) | A.10.2 Allocating responsibilities; A.10.3 Suppliers; A.10.4 Customers |
Numbering starts at .2 in each group because the first item in each section is the objective itself. That quirk matters when you build your Statement of Applicability, because auditors expect your control references to match the standard exactly.
What do auditors look for in each control group?
Auditors want to see that each applicable control is designed for your AI context, implemented and producing evidence. The summaries below describe the intent of each group and typical evidence.
A.2 and A.3: policy and organisation
An approved AI policy that fits with your information security, privacy and quality policies, reviewed at planned intervals. Defined AI roles and a channel for staff to report concerns about AI systems. Evidence: policy with approval record, RACI, and a reporting procedure that people know about.
A.4: resources
Documentation of the data, tools, computing resources and people each AI system relies on. Evidence: an inventory for each AI system listing datasets, models, platforms, infrastructure and the competences of the team.
A.5: impact assessment
A defined process to assess the potential consequences of AI systems for individuals, groups and society, with documented results. This is one of the most distinctive parts of ISO 42001, because it looks beyond risks to the organisation.
Common mistake: Treating an AI system impact assessment as a copy of your DPIA. A DPIA focuses on personal data. An AI impact assessment must also consider issues such as fairness, safety, accessibility and wider societal effects, even where little personal data is involved.
A.6: life cycle
The largest group, with nine controls. It covers responsible development objectives and processes, requirements, design documentation, verification and validation, deployment, operation and monitoring, technical documentation and event logs. Evidence: design records, test results against acceptance criteria, release approvals, monitoring dashboards and log retention settings.
A.7: data
Controls on how data for AI is acquired, prepared, quality-checked and traced. Evidence: data sheets, provenance records, quality metrics and documented preparation steps such as labelling and cleaning.
A.8: information for interested parties
Users and other interested parties get the information they need, incidents are communicated, and external reporting obligations are met. Evidence: user documentation, transparency notices and an incident communication procedure.
A.9: use
Processes and objectives for responsible use, and use of each system in line with its intended purpose. This group matters most for organisations that deploy AI built by others.
A.10: third parties and customers
Responsibilities are allocated across the AI supply chain, suppliers are managed, and customer needs are considered. Evidence: contracts, supplier assessments and customer-facing documentation. A vendor and third-party risk management process is the natural home for these controls.
How do ISO 42001 Annex A controls map to other frameworks?
Many organisations already hold ISO/IEC 27001 or are preparing for the EU AI Act. Mapping lets you reuse evidence rather than duplicate it.
| ISO 42001 area | ISO/IEC 27001:2022 | EU AI Act (high-risk) | NIST AI RMF 1.0 |
|---|---|---|---|
| A.2 Policies, A.3 Organisation | Clause 5, A.5.1 Policies, A.5.2 Roles | Quality management system, Art. 17 | Govern |
| A.5 Impact assessment | No direct equivalent | Risk management, Art. 9; fundamental rights impact assessment for certain deployers, Art. 27 | Map |
| A.6 Life cycle | A.8.25 to A.8.29 secure development and testing | Technical documentation, Art. 11; record-keeping, Art. 12; accuracy and robustness, Art. 15 | Measure, Manage |
| A.7 Data | A.5.12 Classification, A.8.10 to A.8.12 data handling | Data and data governance, Art. 10 | Map, Measure |
| A.8 Information for interested parties | A.5.24 to A.5.26 incident management | Transparency, Art. 13; serious incident reporting, Art. 73 | Govern, Manage |
| A.9 Use | A.5.10 Acceptable use | Deployer obligations, Art. 26; human oversight, Art. 14 | Manage |
| A.10 Third parties | A.5.19 to A.5.23 supplier relationships | Value chain responsibilities, Art. 25 | Govern |
Timing matters for the AI Act column. Regulation (EU) 2026/1744, the Digital Omnibus on AI, was published on 24 July 2026 and entered into force on 27 July 2026. It moved the application of high-risk obligations for Annex III systems to 2 December 2027, and for AI embedded in products covered by Annex I legislation to 2 August 2028. Prohibited practices have applied since 2 February 2025 and general-purpose AI obligations since 2 August 2025.
In the US, the NIST AI Risk Management Framework, released on 26 January 2023, is voluntary and organised around Govern, Map, Measure and Manage. It fits well alongside ISO 42001 for organisations that sell into both markets.
Cross-framework mapping is where software saves the most effort. Enactia’s Compliance Universe uses AI to map controls across ISO 42001, ISO 27001, the EU AI Act and other frameworks, so one piece of evidence can support several requirements. To try it with your own AI inventory, start your 14-day free trial.
How do you build a Statement of Applicability for ISO 42001?
List all 38 Annex A controls, mark each as applicable or not, justify the decision, and record implementation status and evidence. The SoA should trace back to your AI risk assessment and impact assessments.
- Define scope: which AI systems, roles (developer, provider, user) and locations are in the AIMS.
- Inventory AI systems: purpose, data, models, suppliers and owners for each system.
- Assess risks and impacts: run the AI risk assessment (clause 6.1.2) and AI system impact assessments (clause 6.1.4 and A.5).
- Select controls: choose treatments, then compare them with Annex A to check for gaps.
- Write justifications: link each included control to a risk, requirement or objective, and explain each exclusion.
- Record status and evidence: implemented, partly implemented or planned, with owners and dates.
- Review: update the SoA when AI systems, suppliers or regulations change.
In practice: A company that only uses third-party AI tools may justify excluding some development controls in A.6, but it will rarely be able to exclude A.9 on responsible use or A.10 on suppliers. Our guide to the ISO 27001 Statement of Applicability covers the same technique for information security.
Supporting tools help here too: a risk management module for AI risks and a policy management workflow for the AI policy and procedures.
How does certification against ISO 42001 work?
Accredited certification bodies audit your AIMS in two stages, then run surveillance audits. ISO/IEC 42006:2025, published in July 2025, sets requirements for bodies that audit and certify AI management systems, including the competence of their auditors. When choosing a certification body, ask about its accreditation and experience with AI systems like yours.
For a step-by-step plan, see our ISO 42001 certification roadmap.
Key takeaways
- ISO 42001 Annex A contains 38 controls in nine objectives, from A.2 AI policies to A.10 third-party and customer relationships.
- Annex A is a reference set: you select controls through risk treatment and justify choices in a Statement of Applicability.
- A.5 impact assessment and A.7 data controls are the areas that differ most from ISO 27001.
- Mapping Annex A to ISO 27001, the EU AI Act and the NIST AI RMF reduces duplicated work.
- High-risk AI Act obligations now apply from 2 December 2027 or 2 August 2028, depending on the system.
Frequently asked questions
How many controls are in ISO 42001 Annex A?
There are 38 controls grouped into nine control objectives, A.2 to A.10. The largest group is A.6 on the AI system life cycle, with nine controls. The smallest is A.3 on internal organisation, with two. Annex B of the standard gives implementation guidance for every control in Annex A.
Do we have to implement all 38 Annex A controls?
No. You determine the controls needed to treat your AI risks, then compare them with Annex A to check nothing necessary is missing. You can exclude controls that are not relevant, but each exclusion must be justified in your Statement of Applicability, and auditors will test whether those justifications are reasonable.
How is ISO 42001 Annex A different from ISO 27001 Annex A?
ISO 27001 Annex A has 93 information security controls. ISO 42001 Annex A has 38 controls focused on AI, including impact assessments on individuals and society, data provenance and quality, AI life cycle documentation and responsible use. Many organisations run both, reusing shared evidence on policies, suppliers and incidents.
Does ISO 42001 certification mean compliance with the EU AI Act?
Not automatically. ISO 42001 supports many AI Act requirements, such as risk management, data governance, documentation and supplier responsibilities, but the Act has specific legal obligations and conformity procedures. Treat ISO 42001 as a strong governance foundation and map it to the Act article by article for your systems.
Which Annex A controls matter most for AI users rather than developers?
Organisations that deploy AI built by others usually focus on A.2 policies, A.3 roles, A.5 impact assessment, A.8 information for interested parties, A.9 responsible and intended use, and A.10 supplier relationships. Some A.6 development controls may be excluded, but monitoring and logging in operation often remain relevant.
Turning Annex A into a working AIMS
ISO 42001 Annex A controls give you a tested structure for governing AI, but their value depends on how well they reflect your real systems. Start with a clear AI inventory, run proper impact assessments, and let those drive a Statement of Applicability you can defend.
Enactia’s ISO 42001 compliance tools help you manage the AIMS alongside your other frameworks. To discuss your programme, contact us or book a demo or start your 14-day free trial.
