ISO/IEC 42001:2023 is the first international standard for AI Management Systems (AIMS). Published in December 2023, it provides a systematic framework for organisations that develop, provide, or use AI systems to manage AI-related risks, demonstrate responsible AI governance, and meet growing customer and regulatory demands for structured AI accountability.
This guide explains what ISO 42001 requires, how it works, and what the certification process looks like — with practical guidance for SaaS companies building AI into their products.
What ISO 42001 actually is
ISO 42001 is a management system standard — the same category as ISO 27001 (information security) and ISO 9001 (quality management). It doesn't prescribe specific AI techniques or certify individual AI models. Instead, it certifies that an organisation has a functioning AI Management System: a set of policies, processes, roles, and controls for governing AI throughout its lifecycle.
The standard covers three types of organisations:
- AI providers/developers: organisations that build and deploy AI systems
- AI deployers/operators: organisations that integrate third-party AI into their products or services
- AI users: organisations that use AI internally (HR tools, document processing, analytics)
Most AI-enabled SaaS companies fall into two or all three categories simultaneously — building proprietary AI while also integrating third-party models (GPT-4, Claude, Gemini) and using AI internally.
ISO 42001 vs ISO 27001: what's different
ISO 42001 follows the same High-Level Structure (HLS) as ISO 27001 — so if you're already ISO 27001 certified, the management system framework is familiar. But the subject matter is fundamentally different:
| Dimension | ISO 27001 | ISO 42001 |
|---|---|---|
| Scope | Information security management | AI system lifecycle management |
| Primary risk | CIA triad (confidentiality, integrity, availability) | AI harms (bias, safety, autonomy, transparency) |
| Key assessment | Information security risk assessment | AI risk assessment + AI Impact Assessment (AIIA) |
| Controls framework | Annex A: 93 controls across 4 themes | Annex A: AI-specific controls (A.2–A.10) |
| Certification what | Certifies the ISMS | Certifies the AIMS |
| SoA equivalent | Statement of Applicability (Annex A) | Statement of Applicability (AI Annex A) |
| Overlap | Strong — ISMS already covers IT security | Covers AI ethics, fairness, transparency — not in ISO 27001 |
If you're ISO 27001 certified, you can pursue an integrated ISMS+AIMS certification — sharing many of the management system elements (leadership commitment, internal audit, management review, documented information) while adding AI-specific processes on top.
The ISO 42001 clause-by-clause requirements
Clause 4 — Context of the Organization
Before establishing an AIMS, the organisation must understand its AI-specific context:
- Cl. 4.1: Internal and external factors relevant to AI — including regulatory environment (EU AI Act, GDPR intersection), societal expectations, competitive pressures, and the nature of AI use
- Cl. 4.2: Interested parties — customers, regulators, affected communities, employees, and their requirements regarding AI governance
- Cl. 4.3: AIMS scope — which AI systems, products, and organisational units are in scope. This must be documented and maintained.
- Cl. 4.4: Establishing the AIMS — the management system must be established, implemented, maintained, and continually improved
Clause 5 — Leadership
Top management accountability is a central theme of ISO 42001:
- Cl. 5.1: Leadership and commitment — management must actively champion responsible AI, allocate resources, and hold themselves accountable
- Cl. 5.2: AI Policy — a formal AI Policy must be established that commits to responsible AI principles, is communicated internally, and is available to interested parties (customers, regulators)
- Cl. 5.3: Roles and responsibilities — clear ownership of AI governance: who owns AI risk, who conducts impact assessments, who oversees compliance
What an AI Policy must cover: responsible use commitments, AI governance principles, human oversight expectations, and how the AI Policy interacts with the organisation's broader information security and privacy policies.
Clause 6 — Planning
The planning requirements are where most organisations have the biggest gaps:
- Cl. 6.1: AI risk assessment — identify risks from AI systems (technical, social, operational), analyse them, evaluate against acceptance criteria, and select treatment options
- Cl. 6.1.3: AI risk treatment plan with Annex A controls — select applicable Annex A controls and produce a Statement of Applicability
- Cl. 6.2: AI objectives — measurable objectives for AI governance performance (fairness metrics, incident rates, training completion rates)
The AI risk assessment in ISO 42001 is different from an information security risk assessment. It focuses on AI-specific risks: bias and discrimination, safety failures, autonomy erosion, privacy violations through AI, transparency gaps, and third-party AI dependency risks.
Clause 8 — Operation (the most important clause)
Clause 8 covers the operational controls for AI systems across the full lifecycle:
- Cl. 8.2: AI risk assessment process — actually performing the risk assessment, not just documenting a methodology
- Cl. 8.3 / Annex A.5: AI Impact Assessment (AIIA) — before deploying an AI system, a systematic assessment of potential harms to individuals and society must be conducted. This covers: fairness impacts, safety considerations, privacy implications, autonomy effects, and social impact
- Cl. 8.4 / Annex A.6: AI system lifecycle controls — controls across design, development, testing, deployment, monitoring, and decommissioning
- Cl. 8.5 / Annex A.10: Third-party AI management — due diligence for AI providers, contractual requirements, ongoing oversight
AI Impact Assessment (AIIA): what it is and when it's required
The AI Impact Assessment is one of ISO 42001's most distinctive requirements. It's not a Data Protection Impact Assessment (DPIA) — it's broader, covering harms beyond privacy:
| AIIA Dimension | What to assess | Example questions |
|---|---|---|
| Fairness | Discriminatory outcomes by protected characteristics | Does the model perform differently across demographic groups? |
| Safety | Physical, psychological, financial harm potential | What happens when the model makes a wrong prediction at scale? |
| Privacy | Privacy risks from training data, inferences, outputs | Can the model memorise and reproduce personal data? |
| Autonomy | Erosion of human decision-making or agency | Are users relying on AI in ways that reduce their own judgment? |
| Transparency | Ability of affected persons to understand AI decisions | Can a user understand why the AI made a decision affecting them? |
| Social impact | Broader societal effects of the AI system at scale | What happens to society if this AI system operates at 10× scale? |
An AIIA should be conducted before initial deployment and repeated when the AI system changes significantly, when new use cases are identified, or when the deployment context changes.
ISO 42001 Annex A controls overview
Annex A contains AI-specific controls across 9 areas (A.2–A.10). Unlike ISO 27001 Annex A, these are not prescriptive technical controls — they are governance, process, and accountability controls:
- A.2: Policies related to AI — what the AI policy must contain
- A.3: Internal organisation — roles, responsibilities, governance structure for AI
- A.4: Resources for AI systems — ensuring adequate compute, data, and human expertise
- A.5: Assessing impacts of AI systems — AIIA methodology and documentation
- A.6: AI system lifecycle — controls across design, development, deployment, monitoring, decommissioning
- A.7: Data for AI systems — data quality, provenance, bias assessment, labelling
- A.8: Information for interested parties — transparency disclosures, explainability
- A.9: Use of AI systems — responsible use guidance for operators and consumers
- A.10: Third-party and customer relationships — supplier due diligence, contractual requirements
ISO 42001 and the EU AI Act: how they align
ISO 42001 is not the same as EU AI Act compliance — but it's a strong foundation for it. The key intersections:
- Article 9 (Risk management for high-risk AI): ISO 42001 Clauses 6.1–6.2 and 8.2–8.3 directly map to Art. 9 requirements for a risk management system
- Article 17 (Quality management for high-risk AI): ISO 42001's AIMS structure substantially satisfies Art. 17 quality management requirements
- Article 10 (Data governance): ISO 42001 Annex A.7 data governance controls align with Art. 10 data requirements for high-risk AI
- Article 14 (Human oversight): ISO 42001 Annex A.3 and A.6 cover human oversight mechanisms
- Article 50 (Transparency): ISO 42001 Annex A.8 transparency controls support Art. 50 disclosure requirements
An organisation with ISO 42001 certification is not automatically EU AI Act compliant — the Act has specific conformity assessment, registration, and technical documentation requirements that go beyond the management system. But an AIMS dramatically reduces the effort to achieve EU AI Act compliance for high-risk AI providers.
The ISO 42001 certification path
Certification follows the same pattern as ISO 27001:
- Gap assessment: Identify current state against ISO 42001 clauses and Annex A
- AIMS implementation: AI Policy, AI risk assessment, AIIA methodology, lifecycle controls, data governance, transparency procedures — typically 3–6 months for a 50-person AI company
- Internal audit: Conduct an internal audit of the AIMS before Stage 1
- Stage 1 audit: Certification body reviews documentation — AIMS scope, AI Policy, risk assessment, SoA, documented procedures
- Stage 2 audit: Evidence that controls are implemented and operating — AI impact assessments, lifecycle controls in practice, training records, transparency disclosures
- Certificate issued: Valid 3 years, subject to annual surveillance audits
For companies already ISO 27001 certified, pursuing an integrated ISO 27001 + ISO 42001 certification is the most efficient path — sharing the management system infrastructure while adding AI-specific layers.
Start with our free ISO 42001 AI Management System Gap Assessment to understand your current readiness. For EU AI Act compliance, see the EU AI Act Compliance Checklist and AI Risk Register Generator. For the underlying information security management system, see the ISO 27001 Gap Assessment.