ISO 42001 Certification Process

Stefan Coetzeecreated 2026-05-12updated 2026-06-086 min readcross-cutting
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.

End-to-end mechanics for ISO/IEC 42001:2023 certification. Same Annex SL mechanics as ISO 27001 — Stage 1 + Stage 2 audits, annual surveillance, three-year recertification. Audit-practice maturity is still early (certification bodies started offering audits late 2024 / early 2025), so this atom emphasizes current state and emerging conventions rather than settled practice.

Implementation timeline (first-time certification)

For an org without prior ISO management-system certification:

  • Months 0-2: scope decision, gap analysis against ISO 42001:2023, methodology selection.
  • Months 2-6: AI policy development, AI risk methodology, AI system inventory, role assignment, impact assessment process design, control implementation work.
  • Months 6-9: initial AI risk assessment, AI system impact assessments, SoA production, control implementation completion, training rollout, evidence base buildup.
  • Months 9-10: internal audit dry run.
  • Month 10: management review.
  • Months 10-11: Stage 1 audit.
  • Months 11-13: Stage 2 audit. Certification decision.

For an org already certified to ISO 27001:

  • Implementation overlap reduces timeline by 30-50%.
  • Combined audit option compresses external-audit timeline.

Realistic ranges: 6-12 months for ISO 27001-mature orgs adding ISO 42001, 12-18 months for orgs implementing both simultaneously without prior certifications.

The four-audit cycle

Identical to ISO 27001:

  • Year 1: Stage 1 + Stage 2 = initial certification.
  • Year 2: Surveillance audit 1.
  • Year 3: Surveillance audit 2.
  • Year 4: Recertification audit.

Findings categorized as major nonconformity (blocks certification), minor nonconformity (must be addressed within agreed window), observation / opportunity (non-binding feedback).

Stage 1 audit

Documentation and readiness review. Auditor checks:

  • AIMS scope documented.
  • AI policy in place.
  • AI risk methodology defined.
  • AI risk register populated.
  • SoA produced with Annex A coverage.
  • AI system inventory in place.
  • AI system impact assessments performed for in-scope systems.
  • Internal audit programme planned.
  • At least one management review held.
  • Basic control implementations evidenced.

Outcome: ready for Stage 2 or recommendations for further work.

Stage 2 audit

Full implementation audit:

  • Document review (continued).
  • Interviews with top management, AI leadership, AI system owners, AI risk owners, control owners, sampled staff.
  • Evidence sampling: AI policy communication evidence, training records, impact assessment records, risk treatment records, vendor management records, audit logs, incident records.
  • Walkthroughs of critical processes: impact assessment process, risk treatment process, change management for AI systems, incident response for AI-related incidents.
  • AI system reviews: sample of in-scope AI systems with full lifecycle evidence.

Outcome: certification decision. Major nonconformities block certification until remediated.

Audit-practice maturity (current state, 2026)

Certification bodies started offering ISO 42001 audits in late 2024 / early 2025. Current maturity:

  • First-year auditors: trained in ISO 42001 fundamentals and Annex A. Variable AI-domain experience.
  • AI-domain depth: variable. Auditors with prior ML / data-science / AI-engineering background bring depth; others rely on the framework structure.
  • Common findings: AI policy generic, impact assessment treated as one-shot, AI system inventory incomplete, vendor relationship evidence thin, lifecycle-stage tracking missing.
  • Combined audits: ISO 27001 + ISO 42001 combined audits common; auditors handle both standards in single engagement.
  • Inter-auditor variance: high in early state. Expected to converge as audit-practice matures through 2025-2027.

Implication for buyers: an ISO 42001 certificate issued in 2025-2026 carries less convergent meaning than a mature ISO 27001 certificate. Buyers asking for the SoA and the audit report (where supplier willing to share bilaterally) get more substance than the certificate alone.

Implication for implementers: choosing a certification body with strong AI-domain auditing capability matters more in early state than it will in 2028+ when audit practice has matured.

Accreditation chain

Standard ISO mechanics:

  • IAF coordinator.
  • Accreditation bodies (UKAS, DAkkS, ANAB, others) accredit certification bodies under ISO/IEC 17021-1.
  • ISO/IEC 42006 (in development) — will formalize the AIMS-specific certification body requirements. Until 42006 publishes, accreditation under 17021-1 is the baseline.

Buyer-side verification: certificate names the certification body; accreditation status verifiable via the accreditation body. Unaccredited certificates exist and are detectable in buyer due diligence.

Certification body selection

Selection criteria for ISO 42001 specifically:

  • Accreditation confirmation for the AIMS scope. Confirm via accreditation body website.
  • AI-domain audit experience: ask about auditor backgrounds. Senior auditors with ML / data-science / AI-engineering background are scarce; firms with AI-domain leads in their audit practice carry depth.
  • Combined-cert capability: ISO 27001 + ISO 42001 combined audits reduce overhead. Some bodies also offer ISO 27701 combined for three-way.
  • Sector experience: AI-feature SaaS, regulated industries, AI-product companies all have specific concerns; ask about prior similar customers.
  • EU AI Act preparation: certification bodies tracking the AI Act harmonization process closely will be positioned to handle the harmonized-standard auditing once formal.

Common bodies with early ISO 42001 capability:

  • BSI Group — UK origin; published ISO 42001 readiness content early.
  • DNV — strong in maritime, energy; AIMS capability building.
  • TÜV Süd — German technical inspection; combined offerings with ISO 27001.
  • LRQA — global; combined offerings.
  • Schellman — US-origin; combined SOC 2 + ISO 27001 + ISO 42001 positioning.
  • A-LIGN — US-origin; similar combined.
  • PECB — emerging.

Cost orientation

For a 50-300 person AI-feature SaaS company:

  • First-time ISO 42001 alone: €10-25k for Stage 1 + Stage 2 + first surveillance. Plus consultant support if used. Plus 4-8 person-months of internal effort.
  • Combined ISO 27001 + ISO 42001 first-time: €25-50k vs ~€30-65k separate. Combined-audit savings real.
  • Annual surveillance ISO 42001: €4-8k each.
  • Recertification ISO 42001: €10-20k.

Costs likely to rise as audit-practice matures and demand outpaces auditor supply through 2026-2027.

Common procedural pitfalls

  • Treating ISO 42001 as ISO 27001 plus Annex A bolt-on. Impact assessment, lifecycle, vendor relationships need genuine treatment.
  • AI system inventory underestimated. Often missed early; foundational to risk and impact work.
  • Generic AI policy. Boilerplate "trustworthy AI" language without org-specific commitments.
  • Impact assessment as one-shot. Cl 8.4 expects ongoing impact assessment, especially at significant changes.
  • Vendor evidence thin. A.10 third-party controls expect substantive due diligence and ongoing monitoring of model providers.
  • Selecting auditor without AI-domain capability check. Early-stage auditor variance produces uneven audit experiences. Worth asking about auditor backgrounds.
  • EU AI Act timing misalignment. Some orgs pursuing ISO 42001 before AI Act enforcement applies to them; useful procurement signal but may not directly satisfy AI Act compliance until harmonization completes.

Path for small / solo operations

Same considerations as ISO 27001 cost-gating:

  • Small AI-product companies may struggle with ISO 42001 cost. Cyber Essentials Plus equivalent for AI doesn't exist yet.
  • NIST AI RMF alignment (voluntary, non-certified) is the lighter alternative for orgs not yet warranting cert.
  • Combined cert when revenue justifies — typically when AI-feature procurement signal exceeds cert cost with margin.
  • For solo / small-team consulting: cert not warranted. Implement aligned to ISO 42001 structure, maintain evidence, present client-by-client when engagement requires.

SRE and AI-agent fit summary

The certification process is largely standard ISO mechanics; the AI-specific concerns are in implementation and evidence content rather than process mechanics. SRE / AI-agent specific concerns:

  • AI system inventory must include agents, models, prompts, tool definitions, retrieval corpora.
  • Impact assessment must address agent-specific impacts.
  • Lifecycle tracking must accommodate the iteration speed of agent development.
  • Vendor monitoring must include detection of model-behavior changes from providers.

Stefan-context implementation sketch

  • Solo / small-team operation: certification not currently warranted.
  • For client engagements: implement aligned to ISO 42001 structure, document evidence in vault.
  • Track EU AI Act harmonization progress; certification timing benefits from harmonization completion.
  • For procurement-readiness: combined ISO 27001 + ISO 42001 alignment is the positioning to maintain.

See also