ISO 42001

Stefan Coetzeecreated 2026-05-12updated 2026-06-085 min readMOC
Vault note, not reviewed against the source. Written in the knowledge vault on 2026-05-12 by models working with Stefan Coetzee and published as it stands, with private addresses, e-mail addresses and an employer name redacted. Check claims against the primary source before relying on them.

Map of ISO/IEC 42001:2023 as the international standard for AI Management Systems (AIMS). The 27001-equivalent for AI governance: Annex SL-aligned management-system requirements (Cl 4-10) plus Annex A controls specific to AI system development, deployment, and operation. Reference cluster for AI-agent system compliance, model-vendor relationships, autonomous-operations governance, and the procurement signals appearing with EU AI Act 2026-2027 enforcement.

Anchors

  • position: current view, dated, revisable
  • anchors: primary documents, ISO/IEC JTC 1/SC 42, certification bodies, named voices

Provenance

  • ISO/IEC JTC 1/SC 42 — joint technical committee subcommittee on Artificial Intelligence. Established 2017. Working groups on foundational standards, computational approaches, trustworthiness, use cases and applications, governance.
  • ISO/IEC 42001:2023 — published 18 December 2023. The first AI-specific management system standard. Result of multi-year SC 42 work; first major-ISO AI governance publication.
  • Companion standards (SC 42 family):
    • ISO/IEC 22989:2022 — AI concepts and terminology
    • ISO/IEC 23053:2022 — Framework for AI using ML
    • ISO/IEC 23894:2023 — AI risk management guidance
    • ISO/IEC 5338:2023 — AI system life cycle processes
    • ISO/IEC 5339:2024 — AI applications use cases
    • ISO/IEC 27090 (in development) — AI security
    • ISO/IEC 27091 (in development) — AI privacy
    • ISO/IEC 42005 (in development) — AI system impact assessment
    • ISO/IEC 42006 (in development) — Requirements for AI management system audit/certification bodies
  • First certifications issued late 2024 / early 2025 by BSI, DNV, TÜV Süd, others. Audit-practice maturity still ramping as of 2026-05-12.

What ISO 42001 is, in one paragraph

ISO/IEC 42001:2023 is a certifiable management-system standard for AI. It defines requirements for establishing, implementing, maintaining, and continually improving an AI Management System (AIMS). Annex SL aligned (shares high-level structure with ISO 9001, ISO 14001, ISO 27001). Applies to any organization developing, providing, or using AI systems, regardless of size or sector. Annex A contains a reference set of AI-specific controls covering policy, organizational roles, resources, AI system impact assessment, AI system lifecycle management, data handling, transparency, use of AI systems by the org, and third-party relationships. Sister standard to ISO 27001; many implementations integrate them.

Mandatory clauses (Cl 4-10, Annex SL aligned)

Same structure as ISO 27001 and other Annex SL standards:

ClauseTitleAtom
4Context of the OrganizationISO 42001 Clause Structure
5LeadershipISO 42001 Clause Structure
6PlanningISO 42001 Clause Structure
7SupportISO 42001 Clause Structure
8OperationISO 42001 Clause Structure
9Performance EvaluationISO 42001 Clause Structure
10ImprovementISO 42001 Clause Structure

(Consolidated into a single clause-structure atom because the structure mirrors ISO 27001 mechanics; the AI-specific content lives mostly in Annex A.)

Detail in ISO 42001 Clause Structure.

Annex A controls (AI-specific)

Approximately 38 controls organized around 10 control objectives:

  • A.2 Policies related to AI (~2 controls)
  • A.3 Internal organization (~2 controls)
  • A.4 Resources for AI systems (~5 controls)
  • A.5 Assessing impacts of AI systems (~4 controls)
  • A.6 AI system lifecycle (~7 controls)
  • A.7 Data for AI systems (~4 controls)
  • A.8 Information for interested parties of AI systems (~5 controls)
  • A.9 Use of AI systems (~2 controls)
  • A.10 Third-party and customer relationships (~3 controls)

Detail in ISO 42001 Annex A Controls.

The AI system lifecycle and AI system impact assessment are the load-bearing concepts. Lifecycle includes: planning, design and development, verification and validation, deployment, operation and monitoring, re-evaluation and retirement. Impact assessment is required for AI systems before significant deployment decisions.

Integration with ISO 27001 and ISO 27701

ISO 42001 is sibling to:

  • ISO 27001 (Information Security Management System) — many controls overlap (information security, supplier management, asset management, incident management).
  • ISO 27701 (Privacy Information Management System) — overlap on PII handling, data protection impact assessment, data subject rights.
  • ISO 9001 (Quality Management System) — process discipline overlap.

Combined implementation is the dominant pattern. Detail in ISO 42001 vs ISO 27001 Integration.

Certification process

Standard ISO certification mechanics:

  • Stage 1 audit (documentation review)
  • Stage 2 audit (implementation audit)
  • Initial certification, 3-year validity
  • Annual surveillance audits
  • Recertification at three-year mark

Certification bodies started offering ISO 42001 audits in late 2024. Audit-practice maturity is still developing; auditor depth on AI-specific controls varies materially across providers.

Detail in ISO 42001 Certification Process.

ISO 42001 in the wider regulatory landscape

ISO 42001 connects to multiple regulatory and standardization streams:

  • EU AI Act (Reg 2024/1689) — ISO 42001 is widely expected to become the harmonized standard supporting AI Act compliance for general-purpose and high-risk AI systems. Formal harmonization process underway in 2025-2026.
  • NIST AI RMF (US) — voluntary; outcomes-oriented. Maps to ISO 42001 at a function level (govern / map / measure / manage maps to ISO clauses + controls). Combined implementation is feasible.
  • National AI strategies — various jurisdictions (UK, Canada, Singapore, Japan, Australia, Brazil) referencing ISO 42001 in voluntary or upcoming-regulatory contexts.
  • Sector-specific AI guidance — financial services (BIS, FSB), healthcare (FDA, EMA, MHRA), automotive (UN R155-style extensions) often reference or build on ISO 42001 structure.

Why this matters for SRE and AI-agent work

  • AIMS is the standard frame for AI governance. Like ISO 27001 became the InfoSec governance frame, ISO 42001 is positioning to become the AI governance frame.
  • Procurement signal incoming. As EU AI Act enforcement ramps (Aug 2026, Aug 2027), buyers will increasingly specify AI governance evidence. ISO 42001 cert is the most defensible signal currently available.
  • Lifecycle discipline. The AI system lifecycle concept formalizes what SRE / DevOps teams already do informally for ML systems. Aligns with MLOps practice.
  • Impact assessment as a control. AI impact assessment is mandatory for in-scope AI systems. Covers fairness, robustness, transparency, accountability, environmental impact, societal impact. Formal process required.
  • Third-party / customer relationships clause. Model provider relationships (Anthropic, OpenAI, Google, Cohere, etc.) get explicit treatment. Vendor due diligence, contract terms, ongoing monitoring required.
  • Documentation expectations are heavy. Like ISO 27001, ISO 42001 expects substantial documented information. Lean orgs need to design the documentation to add value rather than ceremony.

Stefan-context relevance

Stefan does SRE / staff-engineer-track work touching:

  • AI-agent operations (OpenCode plugins, Claude Code, MCP servers, vault-based agent systems)
  • Model-vendor relationships across multiple providers
  • Client engagements where AI features are increasingly load-bearing
  • German Mittelstand procurement that will see AI Act enforcement pressure 2026+

Cluster atoms should:

  • Stay implementation-grounded (named clauses, named controls, real evidence artefacts)
  • Make the ISO 27001 / ISO 42001 / NIST AI RMF / OWASP LLM Top 10 / EU AI Act stack-fit explicit
  • Bridge MLOps and AIOps practice to the management-system language
  • Carry procurement-readiness through Stefan's likely engagement paths
  • ISO 27001 — InfoSec management system, sibling
  • TISAX — automotive supplier assurance; AI-handling concerns increasingly relevant
  • ITIL — service management; AI feature service-management overlap
  • TOGAF — enterprise architecture; AI systems as architectural concerns
  • NIST AI RMF — voluntary US framework, complementary
  • OWASP LLM Top 10 — operational threat taxonomy
  • EU AI Act — regulatory driver

Conventions for this cluster

  • Atoms named ISO 42001 <Topic>.md with consistent structure
  • ISO 42001:2023 is the default version reference
  • Cross-link to ISO 27001 cluster atoms when control content overlaps
  • All atoms cite the relevant clause or control reference number explicitly
  • Practical examples drawn from real AI-agent operations work where possible

See also

ISO 42001 Cluster (pillars MOC) · position · anchors · ISO 27001 Cluster · NIST AI RMF Cluster · EU AI Act Cluster