ISO/IEC 42001:2023 | ISO/IEC 42005:2025 | EU AI Act mapping

ISO/IEC 42001: make AI governance a certifiable management system, not a stack of principles

For Taiwanese companies that provide or use AI: build and certify an AI management system (AIMS) under ISO/IEC 42001:2023, and use the 38 Annex A controls to carry the EU AI Act, customer due diligence and internal generative-AI governance.

ISO/IEC 42001 was published on 18 December 2023 as the first certifiable AI management system standard; certification bodies are accredited under ISO/IEC 17021-1 and ISO/IEC 42006 (July 2025), and ISO/IEC 42005 (May 2025) adds the AI system impact assessment method. It applies in any region and at any size — but it is not the EU AI Act: a certificate proves the system is sound, while legal duties still have to be met clause by clause. This page lines the two up and says when to do which.

18 Dec 2023
ISO/IEC 42001:2023 published (JTC 1/SC 42)
38 controls
Annex A, nine control objectives A.2–A.10
May / Jul 2025
ISO/IEC 42005 impact assessment; ISO/IEC 42006 certification-body requirements
≠
A 42001 certificate is not EU AI Act compliance; the Article 17 QMS harmonised standard EN 18286:2026 was cited on 14 Oct 2026 and is designed to layer on 42001

Which companies should implement ISO 42001 now?

ISO 42001 targets the organisation, not a single model. If any of the following applies, implementation pays back clearly; if none does, we will say you can wait.

  • Customers (finance, healthcare, Japanese/European brands, public tenders) require proof of AI governance in due diligence or bids
  • Your product or service contains AI and will enter the EU or another market with AI regulation, and you need a management backbone for the obligations
  • Large internal roll-out of generative AI (support, documents, code) with concerns about data leakage, bias and accountability
  • You already hold ISO 27001/27701 and want to bring AI into the same management system instead of building a second one
  • You provide AI models or APIs to downstream customers and must state responsibilities and technical documentation clearly
  • The board or parent company requires AI risk in ERM and the annual audit

Small organisations that only use off-the-shelf AI tools occasionally, process no sensitive data and face no customer demand can start with a one-page usage policy from 42001 clause 6 and A.9; certification can wait.

Deliverables

AIMS scope statement and interested-party analysis
Clause 4; AI systems in scope, roles (provider/deployer) and exclusions
AI policy and roles
Clause 5; A.2, A.3
AI system inventory
A.6.2; also the index for EU AI Act Article 49 registration and the GDPR RoPA
AI risk assessment method and records
Clauses 6.1.2–6.1.3; Annex C risk sources
AI system impact assessment template and records
6.1.4, A.5; ISO/IEC 42005; its facts and analysis can be reused where an AI Act Article 27 FRIA is required
Life-cycle and data governance procedures
A.6, A.7; mapped to AI Act Articles 9, 10, 15
Technical documentation and logging specification
A.6.2.7–8; mapped to AI Act Articles 11, 12 and Annex IV
42001 × AI Act mapping
One document set answering both requirements
Internal audit and management review records
Clause 9; stage 1/2 preparation

The most cost-effective path to full compliance

The law requires an outcome, not the most expensive method. Below is the lightest lawful path from the official text and our practice, the evidence you can reuse, the steps you can merge, and what must never be cut — cutting it voids the whole compliance.

Lightest lawful path

  • Classify before spending: most internal enterprise AI (support chat, summarisation, recommendations) is not high-risk and carries only Article 50 transparency and Article 4 AI literacy duties, not the full high-risk set.[Basis:AI Act Article 6; Annex III; Articles 4, 50]
  • Annex III areas limited to narrow procedural tasks, improving completed human results, detecting deviations or preparatory work can be excluded from high-risk under Article 6(3), with a documented and registered assessment.[Basis:Articles 6(3), 6(4), 49(2)]
  • Annex III high-risk obligations apply from 2 Dec 2027 and Annex I from 2 Aug 2028; schedule spend accordingly rather than all in 2026, but do Article 50 now since it has applied since 2 Aug 2026.[Basis:Regulation (EU) 2026/1744; Article 113]
  • Most high-risk AI can use the internal-control conformity assessment (Annex VI) without a notified body; only Annex III point 1 biometrics needs third parties in specific cases.[Basis:Article 43(1), (2)]

Evidence you can reuse

  • An ISO/IEC 42001 management system (risk, data governance, human oversight, records) directly carries Articles 9, 10, 12 and 14; one system serves many AI systems.[Basis:ISO/IEC 42001 clauses 6, 8; AI Act Articles 9–14]
  • For products also under the CRA, CRA conformity gives the presumption for Article 15 cybersecurity; no second set.[Basis:CRA Article 12]
  • The Article 27 fundamental rights impact assessment and the GDPR Article 35 DPIA can be one assessment (explicitly allowed by Article 27(4)).[Basis:AI Act Article 27(4)]

Steps you can merge

  • Merge the AI system inventory with the personal-data RoPA and the CRA product list into one asset register shared by three regimes.[Basis:AI Act Article 49; GDPR Article 30]
  • Fold post-market monitoring and serious incident reporting into the existing PSIRT / security incident flow.[Basis:AI Act Articles 72, 73]

Never cut (cutting voids compliance)

  • Article 5 prohibited practices cannot be marketed or used since 2 Feb 2025, fines €35M / 7%, with no transition.[Basis:Article 5; Article 99(3)]
  • Article 50 transparency (interaction notice, synthetic content marking, deepfake disclosure) is a current obligation since 2 Aug 2026.[Basis:Article 50]
  • Non-EU high-risk providers must appoint an authorised representative; high-risk systems must be registered in the EU database.[Basis:Articles 22, 49]
  • The Article 6(3) exclusion requires a written assessment and registration; "we are not high-risk" cannot be asserted orally.[Basis:Articles 6(4), 49(2)]

Cases and evidence

Anonymised engagements under NDA; client names are not disclosed. Only completed work is described; no outcome is guaranteed.

Manufacturing automation enterprise

Established an AI system inventory, completed AI risk classification, built an algorithm review process and human oversight mechanism for high-risk AI applications, and obtained ISO 42001 certification.

See the cases page →

Decision tools

Use the 11 questions to see readiness (each shows the matching AI Act article), then the cost-effective path to decide what to close first. Everything runs in your browser; no data is sent.

ISO/IEC 42001 readiness self-assessment (11 questions)

Answer per clauses 4–10 and the nine Annex A control objectives; the right column shows the matching EU AI Act articles, so you can see which gap also closes a legal obligation.

  1. 01 · Clause 4–6AIMS scope, interested-party needs and AI roles (provider/deployer) defined, with an AI risk assessment and treatment process?
    AI Act mapping: Art 9
  2. 02 · Clause A.2An AI policy approved by top management, aligned with other policies (security, privacy) and periodically reviewed?
    AI Act mapping: Art 17
  3. 03 · Clause A.3AI roles and responsibilities assigned, with a channel for staff to report concerns?
    AI Act mapping: Art 26(2)
  4. 04 · Clause A.4Data, tooling, computing and human resources documented for each AI system?
    AI Act mapping: Art 11, Annex IV
  5. 05 · Clause A.5An AI system impact assessment process (individuals, groups, society) with documented assessments?
    AI Act mapping: Art 27 (FRIA)
  6. 06 · Clause A.6A managed AI system life cycle (requirements, design, V&V, deployment, operation and monitoring, technical documentation, event logs)?
    AI Act mapping: Art 9, 12, 13, 15, Annex IV
  7. 07 · Clause A.7Acquisition, quality, provenance and preparation of training/validation/test data documented and controlled?
    AI Act mapping: Art 10
  8. 08 · Clause A.8System documentation for users, external reporting and incident communication to interested parties?
    AI Act mapping: Art 13, 50, 73
  9. 09 · Clause A.9Processes and objectives for responsible use of AI, with intended use defined per system?
    AI Act mapping: Art 26, 14
  10. 10 · Clause A.10Responsibilities with suppliers (including model/API providers) and customers allocated by contract and process?
    AI Act mapping: Art 25, 26
  11. 11 · Clause 9–10Performance monitoring, internal audit, management review and continual improvement records?
    AI Act mapping: Art 72

Cross-service evidence reuse matrix

This matrix shows which management system evidence can be shared among ISO 27001, ISO 27701 and ISO 42001, and which privacy-, AI- or product-specific obligations must be built independently. CRA, GDPR and the EU AI Act are laws, not management system standards; an ISO certificate cannot replace fulfilment of legal obligations.

Legend:Primary — Primary: core requirement of that standard — must be built.Shared — Shared: if already built for ISO 27001, can be directly extended without rebuilding.Partial — Partial: provides supporting context, but the corresponding legal obligation still needs to be fulfilled separately.
Evidence typeISO 27001ISO 27701:2025ISO 42001:2023CRA (law)
A. Management system core
Governance policies, objectives and KPIs (ISMS / privacy / AI policy)
Note: ISO 27701 and 42001 can extend the 27001…

ISO 27701 and 42001 can extend the 27001 policy with privacy or AI statements rather than rebuilding the whole framework. CRA is a product law with no management-system policy requirement.

PrimarySharedShared—
Risk assessment method and records
Note: ISO 27701 privacy risk and ISO 42001 AI …

ISO 27701 privacy risk and ISO 42001 AI risk can share the 27001 risk framework method, but each has its own risk sources (privacy threats, AI bias etc.) that must be assessed separately.

PrimarySharedShared—
Supplier and third-party management procedure
Note: 27701 adds processor contract clauses (G…

27701 adds processor contract clauses (GDPR Art. 28); 42001 adds AI supplier responsibility allocation (A.10). CRA upstream component obligations (SBOM, notification) are product duties and must be built separately.

PrimarySharedSharedPartial
Internal audit and management review records
PrimarySharedShared—
Document control, record retention, training records
PrimarySharedShared—
Security incident management procedure (including breach notification)
Note: CRA Art. 14 has its own 24/72-hour vulne…

CRA Art. 14 has its own 24/72-hour vulnerability and serious-incident notification to ENISA, separate from ISMS incident management. Existing incident procedures can inform a PSIRT baseline, but CRA-specific requirements must be built on top.

PrimarySharedSharedPartial
B. Privacy-specific (ISO 27701 / GDPR)
PII inventory and data flow mapping (RoPA)
Note: GDPR Art. 30 requires controllers and pr…

GDPR Art. 30 requires controllers and processors to maintain a RoPA; ISO 27701 provides the management framework for building it, but the mandatory fields and maintenance obligation come from the legal text of GDPR.

—Primary——
Data subject rights procedures (access, erasure, objection, etc.)
—Primary——
Privacy impact assessment (DPIA) report
Note: ISO 27701 privacy risk assessment maps c…

ISO 27701 privacy risk assessment maps closely to the GDPR Art. 35 DPIA obligation, but when a DPIA is mandatory, the nine EDPB WP248 criteria and the obligation to consult the supervisory authority (Art. 36) are legal requirements outside ISO 27701.

—Primary——
Cross-border transfer mechanisms (SCC, TIA, supplementary measures)
Note: ISO 27701 requires identification and re…

ISO 27701 requires identification and recording of cross-border transfers, but which SCC module (Commission Decision (EU) 2021/914) to use, the transfer impact assessment (TIA) and supplementary measures are legal determinations under GDPR Arts. 44–49, requiring legal expertise.

—Partial——
C. AI-specific (ISO 42001 / EU AI Act)
AI system inventory (in-house, purchased, SaaS, generative AI)
Note: EU AI Act Art. 49 requires high-risk AI …

EU AI Act Art. 49 requires high-risk AI system registration in the EU database; the AI system inventory is the foundation for that registration, but the registration obligation itself is a legal requirement.

——Primary—
AI risk assessment and AI system impact assessment (ISO 42005)
Note: ISO 42001 A.5 and ISO/IEC 42005 (28 May …

ISO 42001 A.5 and ISO/IEC 42005 (28 May 2025) provide the impact assessment method; EU AI Act Art. 27 fundamental rights impact assessment (FRIA) can reuse the same facts and analysis, but the FRIA itself is a legal obligation, particularly for deployers of high-risk systems.

——Primary—
Data and model life-cycle controls (A.6, A.7)
——Primary—
Third-party AI and generative AI usage governance (A.9, A.10)
——Primary—
D. Product-specific (CRA legal obligations)
Software bill of materials (SBOM) and component vulnerability tracking
Note: CRA Annex I Part II §1 requires manufact…

CRA Annex I Part II §1 requires manufacturers to identify and document components (including open source); SBOM is a product-level obligation with no direct counterpart in a management system; it cannot be substituted by ISO 27001 or 42001 documents.

———Primary
Coordinated vulnerability disclosure policy (CVD) and six-step PSIRT process
Note: Existing ISMS incident management can in…

Existing ISMS incident management can inform the PSIRT design, but CRA Annex I Part II specific requirements (disclosure channel, obligation to notify component maintainers, 14-day final report) must be built independently.

———Primary
24/72-hour notification mechanism to ENISA and drills
Note: CRA Art. 14 §2, 4; applicable from 11 Se…

CRA Art. 14 §2, 4; applicable from 11 Sep 2026 (including already-marketed products). Taiwanese manufacturers without a principal EU establishment notify per Art. 14 §7 to the CSIRT of the country of the authorised representative or the importer.

———Primary
Annex VII technical documentation and EU declaration of conformity
Note: CRA Annex VII requires documentation of …

CRA Annex VII requires documentation of design, development and maintenance processes, cybersecurity risk assessment and test results, retained for 10 years or the support period (whichever is longer). This is product conformity evidence, qualitatively different from management system documentation.

———Primary

Important: an ISO 27701 certificate does not equal GDPR compliance; an ISO 42001 certificate does not equal EU AI Act compliance; no ISO certificate equals passing CRA conformity assessment. Standards establish a management system; laws require fulfilment of specific statutory obligations. The two must not be conflated.

Procurement evidence

Why Winners belongs on an ISO 42001 consulting shortlist

The fit is strongest when you need an operating system for AI inventory, risk and impact assessment, third-party AI governance, human oversight, model change and incidents. Winners keeps ISO/IEC 42001 separate from EU AI Act legal obligations rather than treating a management-system certificate as legal compliance.

Publicly checkable evidence

  • • Vice-president and director-level consultants lead engagements end to end without outsourcing; published pages state ISO lead-auditor credentials.
  • • Completed ISO 42001 work is described only on an anonymised basis; no fabricated named ISO 42001 client is claimed.
  • • Public buyer guides cover scope, risk/impact/acceptance, third-party AI, human oversight, model changes and post-deployment incidents.

Claims we do not make

  • • ISO 42001 certification is not presented as EU AI Act compliance.
  • • Winners is the implementation consultant, not the certification body.

Our qualifications and approach

How to compare ISO 42001 consultants

CriterionWinners approachWhat to require when buying
AI system inventoryInventories in-house, purchased, SaaS, generative AI and departmental use casesNot just model names — needs use purpose, data, owner, supplier and risk
AI risk and impact assessmentRisk and AI system impact managed separatelyMust have method, approval and re-assessment triggers
Third-party / generative AICovers input data, output review, supplier responsibility and change riskPurchasing a model is not a liability waiver
Sharing with 27001 / 27701Shared document control, internal audit, management review and some supplier processesAI-specific evidence cannot be replaced by security documents
EU AI ActUse-case and role determination; avoids treating a certificate as complianceStandard and law mapped separately
Certification preparationBuilds real operation records; does not back-fill fake recordsConsultant and certification body must be independent
Ongoing governanceAhura stores AI inventory, risk, controls, evidence and approvalsShould still be updatable after project closes

Confirm the consultant and certification body are different organisations before beginning ISO 42001 certification (ISO/IEC 42006 requirement).

Four scenarios most suited to implementing ISO 42001

  • Companies already using ChatGPT, Gemini or Copilot at scale but with no formal AI governance
  • Companies with ISO 27001 / 27701 that want to extend their existing system to cover AI
  • Companies developing AI products, smart manufacturing, healthcare AI or HR AI that need to manage high-risk uses
  • Companies whose customers or supply chains are starting to require AI governance evidence or ISO 42001 certification
Management system × regulation × technology
Built on the ISO/IEC 42001 harmonised structure, sharing the backbone with 27001/27701; the AI Act mapping is maintained by the same team; we run our own AI content pipeline and licence monitoring, so the controls we coach are ones we execute ourselves.
Only controls that certification bodies and regulators accept
Each control maps to an Annex A number and an AI Act article; we do not recommend tools that will not be credited or policies that will not be followed.
Division of work with counsel and certification bodies
We build the system and the evidence chain; legal opinions come from partner lawyers; certification is performed by bodies accredited under ISO/IEC 17021-1 and ISO/IEC 42006 — we accompany, we do not audit for them.
Partner network
Chinese Software Service Industry Association, ITRI, III, ISA Taiwan Section.

Learn More About AI Governance

Certification services × risk glossary × latest insights

💡Latest InsightsView all →

No articles yet

Next steps

Book a consultant (free first diagnosis)

60-minute online meeting on your AI system list, customer requirements and current systems: whether to do 42001, the AI Act or both, and the three things to close first.

Book a consultant

Request scope and timeline

Four-phase scope and schedule based on the number of AI systems, existing 27001 and EU market entry.

Request scope

Entering the EU? See the EU AI Act page

Classification engine, application timeline and three deliverable templates.

See EU AI Act service

Defined terms and framework elements in this service (Risk Glossary)

Each term is taken from the original legal or framework text, with its article or clause; Winners Consulting addresses each one in this service.

Why do Taiwanese companies choose Winners Consulting for ISO 42001?

This service by role

The same service raises different questions for a CEO, CISO, CRO, CFO, CTO, product manager, general counsel and CHRO. Pick your role to see what this service means for your responsibilities.

Ask a consultant on LINE

No forms: the button opens LINE with the page you are reading filled in. Send it and a Winners Consulting consultant will reply. The first diagnosis is free.

Build your ISO/IEC 42001 AI Management System Consulting RFP in three minutes

Answer 7 multiple-choice questions and get an RFP draft you can hand to your manager or send to consultants for quotes.

Question 1 of 7

Why are you looking at ISO/IEC 42001 AI Management System Consulting now?

Winners Consulting track record

Winners Consulting Services is a Taiwan RegTech platform provider and enterprise risk management consultancy.

Trusted long-term by 20+ companies, we work as an industry-focused, long-term partner to our clients.

Named client engagements, published with client permission:

See case studies