Framework · AI governance · Certifiable

ISO 42001 AI management system: controls and certification

ISO/IEC 42001 is the international standard for an AI management system (AIMS). It sets requirements in clauses 4 to 10, including AI risk and impact assessment, and lists 38 reference controls in Annex A. Organisations that develop, provide or use AI can be certified by an accredited certification body.

Key takeaways

What you need to know about ISO 42001

First of its kindThe first certifiable standard for managing AI responsibly.
Impact, not only riskYou assess effects on people and society, not only on the business.
38 controlsAnnex A covers policy, data, life cycle, users and suppliers.
Any AI roleIt applies whether you build, provide or use AI systems.
EU AI ActIt supports AI Act readiness but is not proof of conformity on its own.

Who needs ISO 42001?

No law requires it, but these situations make it valuable.

AI product companiesEnterprise buyers ask how models, data and risks are governed.
Companies using AI in decisionsHR, credit, insurance and health use cases attract scrutiny.
Providers under the EU AI ActA management system helps meet quality and documentation duties.
ISO 27001 holdersExtends an existing ISMS to AI with shared clauses and audits.
Public-sector AI suppliersTenders increasingly ask for AI governance evidence.

QULDEX helps here first: the readiness check shows where you stand before you commit budget.

Clauses 4–10

What does ISO 42001 require?

ISO 42001 requires an AI management system: define your role and AI systems in scope, assess AI risks and impacts, apply controls, and keep checking that it works. Clauses 4 to 10 follow the same structure as ISO 27001.

Clause 4 Plan

Context

Your AI role, interested parties, scope and AI systems.

In QULDEX: AI system inventory set once
Clause 5 Plan

Leadership

AI policy, roles and top-management commitment.

In QULDEX: Policy approvals with sign-off history
Clause 6 Plan

Planning

AI risk assessment, treatment, impact assessment, objectives.

In QULDEX: Risk register generates the SoA
Clause 7 Plan

Support

Resources, competence, awareness, documented information.

In QULDEX: Competence records in the evidence vault
Clause 8 Do

Operation

Run risk treatment and impact assessments in practice.

In QULDEX: Impact assessments per AI system
Clause 9 Check

Evaluation

Monitoring, internal audit, management review.

In QULDEX: Audit workspace and review packs
Clause 10 Act

Improvement

Nonconformity, corrective action, continual improvement.

In QULDEX: Findings flow into CAPA
Annex A · 38 controls

What are the ISO 42001 Annex A controls?

Annex A of ISO/IEC 42001:2023 lists 38 controls in 9 control objectives, from AI policy to suppliers and customers. You choose the ones your risk and impact assessments need. Select any control to see what it asks for and how it maps to the EU AI Act.

38Annex A controls
9control objectives
6.1.4AI impact assessment clause
Dec 2023published

Showing up to 6 per group. Search, filter, or open a group to see all 38.

A.2 Policies 3

  1. A.2.2AI policy

    Document a policy for the development or use of AI systems.

    In QULDEXPolicy approvals with named owners and sign-off history.

    Typical evidence
    Approved AI policy
    Maps to
    EU AI Act Art. 17NIST AI RMF GOVERN

  2. A.2.3Alignment with other organisational policies

    Identify where other policies (security, privacy, quality) apply to AI.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  3. A.2.4Review of the AI policy

    Review the AI policy at planned intervals or after significant change.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Policy review records

A.3 Internal organisation 2

  1. A.3.2AI roles and responsibilities

    Define and allocate roles and responsibilities for AI.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    RACI for AI systems
    Maps to
    NIST AI RMF GOVERN

  2. A.3.3Reporting of concerns

    Give people a way to report concerns about AI systems.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Concern reporting channel and log

A.4 Resources 5

  1. A.4.2Resource documentation

    Identify and document the resources each AI system needs.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  2. A.4.3Data resources

    Document the data resources used by AI systems.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Data inventory per AI system

  3. A.4.4Tooling resources

    Document tools used for AI development and operation.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  4. A.4.5System and computing resources

    Document system and computing resources for AI systems.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  5. A.4.6Human resources

    Document the competences needed for AI work.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Competence matrix

A.5 Impact assessment 4

  1. A.5.2AI system impact assessment process

    Set a process to assess the impacts of AI systems on individuals, groups and society.

    In QULDEXImpact assessments are templated, versioned and linked to each AI system.

    Typical evidence
    Impact assessment procedure
    Maps to
    EU AI Act Art. 27NIST AI RMF MAP

  2. A.5.3Documentation of impact assessments

    Document impact assessment results and keep them.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Completed impact assessments

  3. A.5.4Impact on individuals or groups

    Assess potential impacts on individuals or groups across the system's life cycle.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Maps to
    EU AI Act Art. 27

  4. A.5.5Societal impacts

    Assess potential societal impacts of AI systems.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

A.6 AI life cycle 9

  1. A.6.1.2Objectives for responsible development

    Set objectives for responsible AI development and integrate them into the life cycle.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  2. A.6.1.3Responsible design and development processes

    Define processes for responsible AI system design and development.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  3. A.6.2.2AI system requirements and specification

    Specify requirements for new AI systems or material changes.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Requirement specifications

  4. A.6.2.3Design and development documentation

    Document design and development based on objectives and requirements.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Design records
    Maps to
    EU AI Act Art. 11

  5. A.6.2.4Verification and validation

    Define and run verification and validation measures.

    In QULDEXTest results attach as evidence per model version.

    Typical evidence
    Validation reports, test results
    Maps to
    EU AI Act Art. 15NIST AI RMF MEASURE

  6. A.6.2.5Deployment

    Plan deployment and confirm requirements are met first.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Deployment approvals

  7. A.6.2.6Operation and monitoring

    Monitor AI systems in operation, including performance and drift.

    In QULDEXMonitoring reports are collected on schedule; issues become CAPA items.

    Typical evidence
    Monitoring dashboards
    Maps to
    EU AI Act Art. 72NIST AI RMF MANAGE

  8. A.6.2.7Technical documentation

    Keep technical documentation for each user group.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Model cards, technical file
    Maps to
    EU AI Act Art. 11

  9. A.6.2.8Recording of event logs

    Decide when to record event logs and keep them.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Logging configuration
    Maps to
    EU AI Act Art. 12

A.7 Data 5

  1. A.7.2Data for development and enhancement

    Define data management processes for developing AI systems.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Maps to
    EU AI Act Art. 10

  2. A.7.3Acquisition of data

    Document how data is acquired and selected.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Data sourcing records

  3. A.7.4Quality of data

    Define and check data quality requirements.

    In QULDEXData quality checks attach as evidence per dataset version.

    Typical evidence
    Data quality reports
    Maps to
    EU AI Act Art. 10

  4. A.7.5Data provenance

    Record the provenance of data across its life cycle.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Provenance records

  5. A.7.6Data preparation

    Define criteria for selecting and preparing data.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

A.8 Information for interested parties 4

  1. A.8.2Documentation and information for users

    Give users the information they need about the AI system.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    User documentation
    Maps to
    EU AI Act Art. 13

  2. A.8.3External reporting

    Give interested parties a way to report adverse impacts.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  3. A.8.4Communication of incidents

    Plan how to communicate AI incidents to users.

    In QULDEXAI incidents are logged and communications tracked in CAPA.

    Typical evidence
    Incident communication plan
    Maps to
    EU AI Act Art. 73

  4. A.8.5Information for interested parties

    Document obligations to report information about AI systems.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

A.9 Use of AI 3

  1. A.9.2Processes for responsible use

    Define processes for the responsible use of AI systems.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Maps to
    EU AI Act Art. 14

  2. A.9.3Objectives for responsible use

    Set objectives to guide responsible use.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  3. A.9.4Intended use

    Make sure AI systems are used as intended and documented.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

    Typical evidence
    Intended-use statements

A.10 Third parties 3

  1. A.10.2Allocating responsibilities

    Allocate responsibilities between you, partners, suppliers and customers.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

  2. A.10.3Suppliers

    Make sure supplier services, products and materials fit your responsible-AI approach.

    In QULDEXEach AI supplier gets one record with assessment and contract terms.

    Typical evidence
    Supplier assessments for models and data

  3. A.10.4Customers

    Consider customer expectations and needs in your responsible-AI approach.

    In QULDEXQULDEX gives this control an owner, a review cycle and an evidence request, and links it to each AI system it applies to.

Control numbers and titles follow ISO/IEC 42001:2023 Annex A. Summaries are QULDEX paraphrases; the standard's text is the authority.

Certification path

How do you get ISO 42001 certified?

Inventory your AI systems, assess risks and impacts, run the controls, then an accredited certification body audits you in two stages. Most organisations take 6 to 12 months; less with an existing ISO 27001 ISMS.

  1. Define your AI role and inventory AI systems

    Decide whether you develop, provide or use AI, and list every AI system in scope with owners.

  2. Assess AI risks and impacts

    Run AI risk assessments and impact assessments on individuals, groups and society; write the Statement of Applicability.

  3. Implement controls and collect evidence

    Put Annex A controls in place across data, life cycle, users and suppliers, and keep the records.

  4. Run an internal audit and management review

    Clause 9 must operate at least once before certification.

  5. Stage 1 audit: design and documents

    The certification body checks scope, AI inventory, impact assessments and documents.

    1–3 days
  6. Stage 2 audit: controls in operation

    Auditors sample evidence, such as validation results and monitoring, to confirm controls work.

    Audit days set by scopeMulti-auditor workspace →
  7. Keep it running for three years

    Year 1Surveillance audit
    Year 2Surveillance audit
    Year 3Recertification audit

Durations are typical planning ranges from QULDEX audit delivery, not requirements of the standard.

Documented information

Which documents does ISO 42001 require?

ISO 42001 names the documented information you must keep. Auditors ask for these first in Stage 1.

Document or recordClauseWhere it lives in QULDEX
AIMS scope4.3Program setup
AI policy5.2Evidence vault · policy library
AI risk assessment process6.1.2Risk register
AI risk treatment and Statement of Applicability6.1.3Control library (generated)
AI system impact assessment process6.1.4Risk register
AI objectives6.2Program dashboard
Evidence of competence7.2Evidence vault
Operational planning and control records8.1Control workflows
Risk assessment and treatment results8.2 · 8.3Risk register
Impact assessment results8.4Risk register
Monitoring results9.1Readiness reports
Internal audit and management review9.2 · 9.3Audit workspace
Nonconformities and corrective actions10.2CAPA automation

Annex A controls add their own records where your SoA includes them.

Time and cost

How long does ISO 42001 take and what drives the cost?

First certification usually takes 6 to 12 months. The number of AI systems and how high-impact they are drive the effort more than headcount alone.

Where the time goes

AI inventory and role2–4 wk
Risk and impact assessment4–8 wk
Implement controls2–5 mo
Internal audit, review3–6 wk
Stage 1 + Stage 24–8 wk

Bars show the upper end of each range on one scale (5 months = full width).

What changes the effort

  • AI systems in scope: each one needs an impact assessment and life-cycle records
  • Your role: developers need more life-cycle controls than users
  • High-impact uses: HR, credit, health and safety uses need deeper assessments
  • Third-party models: supplier controls grow with each model or data provider
  • Existing ISO 27001: shared clauses and audits cut effort
Readiness check · 2 minutes

Are you ready for an ISO 42001 Stage 1 audit?

Stage 1 auditors look for these eight things first. Nothing you enter leaves this page.

4.1Have you decided whether you develop, provide or use AI?
4.3Do you have an inventory of AI systems in scope with owners?
5.2Is there an approved AI policy?
6.1.2Have you assessed AI risks for each system?
6.1.4Have you run impact assessments on individuals and society?
A.7.4Do you check data quality and record data provenance?
A.6.2.6Do you monitor AI systems in operation?
9.2Have you completed an internal audit of the AIMS?
Crosswalk

How ISO 42001 maps to the EU AI Act and NIST AI RMF

ISO 42001 controls line up with many EU AI Act obligations and the NIST AI Risk Management Framework, so one documented control can support several regimes. It supports AI Act readiness but is not, by itself, proof of conformity.

ISO 42001 vs the EU AI Act

ISO 42001EU AI Act
What it isVoluntary, certifiable management-system standardEU regulation (2024/1689) with legal obligations
Who it applies toAny organisation that chooses itProviders, deployers and others placing AI on the EU market
ApproachManagement system for all AI you build or useObligations by risk level: prohibited, high-risk, limited, minimal
OutcomeCertificate from an accredited bodyConformity assessment and CE marking for high-risk AI
TimingAvailable nowObligations phase in from 2025; check current dates
In QULDEXQULDEX maps ISO 42001 controls to AI Act articles so evidence collected once supports both.

AI governance crosswalk

ISO 42001 Annex AEU AI ActNIST AI RMFShared evidence
A.2.2 AI policyArt. 17 (quality management)GOVERNApproved AI policy
A.5.2–A.5.5 Impact assessmentArt. 9, Art. 27MAPImpact assessments
A.7.4 Data qualityArt. 10MEASUREData quality reports
A.6.2.3 · A.6.2.7 DocumentationArt. 11MAPTechnical documentation
A.6.2.8 Event logsArt. 12MANAGELogging configuration
A.8.2 Information for usersArt. 13GOVERNUser documentation
A.9.2 Responsible useArt. 14 (human oversight)MANAGEOversight procedures
A.6.2.4 Verification and validationArt. 15MEASUREValidation reports
A.6.2.6 MonitoringArt. 72 (post-market monitoring)MANAGEMonitoring reports
A.8.4 Incident communicationArt. 73 (serious incidents)MANAGEIncident records

Indicative mapping for planning, not legal advice. The full crosswalk is in the QULDEX control library.

Where QULDEX fits

How QULDEX runs ISO 42001 from AI inventory to certificate

QULDEX is ISO 42001 AI governance and audit management software built from EGV Group's audit delivery, used by the organisation getting certified and by the auditors who test it. Pick your role to see who does what.

For organisations that build, provide or use AI.

  1. InventoryList your AI systemsAI system register with owners and roles
  2. Risk and impactAssess each systemTemplated impact assessments linked to each system
  3. ControlsRun the controls38 controls pre-loaded with owners and evidence requests
  4. Internal auditTest the AIMSAudit workspace, sampling and findings into CAPA
  5. CertificationHost the auditorControlled, logged evidence sharing
  6. SurveillanceKeep it runningLive readiness view per AI system
Without one systemWith QULDEX
AI systems known only to the teams that built themOne AI register with owners and roles
Impact assessments in documents nobody can findVersioned assessments linked to each system
Separate work for ISO 42001 and the AI ActOne control set mapped to both
Model monitoring outside the audit trailMonitoring evidence collected on schedule
10+years of audit delivery500+audits deliveredBoth sidesof the audit on one platformRBACand a full audit trail on every action
FAQ

ISO 42001 questions people ask

What is ISO 42001?

ISO/IEC 42001:2023 is the international standard for an AI management system. It helps organisations develop, provide or use AI responsibly, and it can be certified.

How many controls are in ISO 42001?

Annex A has 38 controls grouped into 9 control objectives. You apply the ones your risk and impact assessments need.

Does ISO 42001 certification mean EU AI Act compliance?

No. ISO 42001 supports many AI Act obligations, such as risk management, documentation and monitoring, but it is not by itself proof of AI Act conformity.

Who can be certified to ISO 42001?

Any organisation that develops, provides or uses AI systems, of any size or sector.

What is an AI system impact assessment?

A documented assessment of how an AI system could affect individuals, groups and society, required by clause 6.1.4 and supported by Annex A.5 controls.

Can ISO 42001 be combined with ISO 27001?

Yes. Both use clauses 4 to 10, so many organisations integrate them with shared risk processes, internal audits and management reviews.

Can QULDEX run ISO 42001 alongside ISO 27001?

Yes. Shared clauses and evidence are mapped once, and each standard keeps its own report.

Sources

References

  1. ISO/IEC 42001:2023, Information technology: Artificial intelligence: Management system. iso.org
  2. Regulation (EU) 2024/1689 (EU Artificial Intelligence Act). eur-lex.europa.eu
  3. NIST AI 100-1, Artificial Intelligence Risk Management Framework (AI RMF 1.0). nist.gov
  4. ISO/IEC 42006, Requirements for bodies providing audit and certification of AI management systems. iso.org
  5. ISO/IEC 27001:2022. iso.org

Reviewed by

Abhishek Yadav

Lead Auditor · QULDEX

Reviewed this page against ISO/IEC 42001:2023: Annex A controls, impact assessment requirements, certification path and the AI governance crosswalk.

Page history
  • : Rebuilt with all 38 Annex A controls, EU AI Act and NIST AI RMF crosswalk and readiness check
  • : Page first published

Find your ISO 42001 gaps in 15 minutes

Answer a short readiness check and get a gap summary by clause and control. No sales call needed to see the result.

Schedule
Book a Demo