Key takeaways
- ISO/IEC 42001 (published December 2023) uses the same harmonized clause structure as ISO 9001, 14001 and 27001, so most of the system already exists in a certified plant.
- The new work is AI-specific: an AI policy, AI risk and impact assessments, life cycle controls, data controls and supplier controls for AI.
- Certification is optional. Customers in regulated supply chains are starting to ask for it, but a documented, aligned system without a certificate is a valid first step.
- Certification bodies now have their own requirements standard, ISO/IEC 42006 (July 2025), which makes audits more consistent.
- Scope tightly: one site, or the AI used in quality and maintenance, is a sensible first scope for a mid-market manufacturer.
ISO/IEC 42001:2023 is the international standard for an AI management system (AIMS). It was published on December 18, 2023, by ISO/IEC JTC 1/SC 42, and it is the only widely recognized AI governance standard you can be certified against. For a manufacturer that already runs ISO 9001, and often ISO/IEC 27001 or ISO 14001, the good news is that most of the machinery is already in place. This guide covers what it asks for, how it plugs into your existing system, and a proportionate path for a mid-market plant.
What ISO/IEC 42001 is
It is a management system standard, not a technical one. It does not tell you how to train a model or which accuracy is good enough. It tells you how to run AI in a controlled, accountable way: set policy, assess risk, implement controls, monitor, audit and improve.
It follows the same harmonized structure as your other ISO systems:
| Clause | Topic | What is new for AI |
|---|---|---|
| 4 | Context of the organization | Your role(s) with AI (provider, user, or both) and the AI systems in scope |
| 5 | Leadership | An AI policy and assigned AI responsibilities |
| 6 | Planning | AI risk assessment, AI risk treatment and AI system impact assessment |
| 7 | Support | Competence and awareness for people working with AI |
| 8 | Operation | Running the assessments and controls during the AI life cycle |
| 9 | Performance evaluation | Monitoring, internal audit, management review |
| 10 | Improvement | Nonconformity and corrective action |
On top of the clauses sit four annexes. Annex A (normative) is a reference set of control objectives and controls. Annex B (normative) gives implementation guidance for those controls. Annex C (informative) lists potential AI-related objectives and risk sources. Annex D (informative) covers use across domains and sectors.
Does it apply to you?
Nobody is legally required to adopt it. It becomes relevant when:
- An automotive, aerospace, medical or electronics customer asks how you govern AI in their parts, and a recognized standard shortens the conversation.
- You are, or may become, a provider of AI under the EU AI Act, for example if you sell machines or products with embedded AI. A management system makes the Act's documentation and quality duties easier to carry.
- You already certify against ISO 9001 or ISO/IEC 27001 and want AI governance to live in the same system, not in a separate binder.
If you only use a couple of low-risk tools, you may not need certification. You may still want to align your AI policy and risk process with it.
The Annex A controls at a high level
Annex A groups its controls under nine themes. In plain language:
- AI policies: a documented AI policy, aligned with your other policies, and reviewed.
- Internal organization: defined roles and responsibilities, and a way for people to raise concerns.
- Resources for AI systems: documented data, tooling, computing and human resources for each system.
- Assessing impacts of AI systems: a process to assess impacts on individuals, groups and society, and to keep records.
- AI system life cycle: objectives for responsible development, requirements, design, verification and validation, deployment, operation and monitoring, technical documentation and event logs.
- Data for AI systems: data quality, provenance, acquisition and preparation.
- Information for interested parties: documentation for users, external reporting, and incident communication.
- Use of AI systems: processes for responsible use, objectives and use according to intended purpose.
- Third-party and customer relationships: allocating responsibilities with suppliers and customers, and supplier controls.
As with ISO/IEC 27001, you produce a Statement of Applicability. Controls that do not apply, for example most development controls if you only buy AI, can be excluded with a justification.
Plant-floor reality: Your biggest AI systems may be ones you never chose. A new filling line arrives with an AI vision module, the MES vendor switches on an "AI assistant", and the maintenance team installs a predictive app from the pump OEM. The third-party controls in Annex A are where a manufacturer does most of its real work.
How it fits with ISO 9001, 27001 and 14001
The shared structure means you extend, not duplicate.
| Element | Reuse from existing system | Add for 42001 |
|---|---|---|
| Document control, records | ISO 9001 | AI-specific records (impact assessments, model versions, test results) |
| Risk methodology | ISO 9001 risk-based thinking, ISO/IEC 27001 risk assessment | AI risk criteria and impact on people, not just on the business |
| Supplier management | ISO 9001 purchasing, ISO/IEC 27001 supplier security | AI-specific clauses: model change notice, test evidence, data use |
| Change control | MOC, engineering change | Retraining, threshold changes and model updates treated as changes |
| Competence and training | ISO 9001 training matrix | AI awareness and role-specific skills |
| Internal audit, management review | Existing program and calendar | AIMS scope added to audit plan and review agenda |
| Incident and nonconformance | CAPA/8D process | AI incidents (bad outputs, drift, misuse) as nonconformance types |
| Environmental aspects | ISO 14001 | Optionally, energy use of AI workloads |
ISO/IEC 27001 covers the security of the information your AI uses. ISO/IEC 42001 covers whether the AI is fit for purpose and used responsibly. If your OT security sits under IEC 62443 instead, see ISO 27001 vs IEC 62443 for how those two divide the work.
What certification involves
- Choose a certification body accredited for ISO/IEC 42001. Since July 2025, ISO/IEC 42006 sets additional requirements for bodies that audit and certify AI management systems, including auditor competence. Ask your current registrar whether they hold accreditation.
- Define scope. Sites, business units and AI systems covered, and your role (provider, user or both).
- Build and operate the system. Policy, inventory, risk and impact assessments, Annex A controls, Statement of Applicability, training, monitoring.
- Generate evidence. At least one internal audit and one management review, plus records showing the controls are working.
- Stage 1 audit: documentation and readiness review.
- Stage 2 audit: on-site or remote evaluation of whether the system is implemented and effective.
- Surveillance audits in the following years, then recertification, on the same cycle as your other ISO certificates.
A proportionate path for the mid-market
You do not have to go straight to certification. A three-step approach works well:
Step 1. Align (months 1 to 3). Build the AI use case inventory, write the AI policy, assign roles, and pick a risk method. The NIST AI RMF is a practical source for the risk work.
Step 2. Operate (months 3 to 9). Run impact assessments on the higher-risk systems, add AI clauses to supplier contracts, add AI changes to MOC, include AI in internal audit and management review.
Step 3. Certify if it pays (months 9 to 12+). Certify when a customer, regulator or your own sales team gives you a reason. Keep the scope narrow, for example one plant or the AI used in quality and maintenance.
Effort and cost
| Starting point | Typical internal effort to certification-ready | Notes |
|---|---|---|
| ISO 9001 + ISO/IEC 27001 certified, narrow scope | 30 to 60 person-days | Mostly AI-specific assessments and supplier work |
| ISO 9001 only | 50 to 90 person-days | Also need information security basics around AI data |
| No certified management system | Consider aligning first | Building a full management system from scratch is a large project |
External costs are the certification audits and any consulting support. Audit days depend on scope and headcount, so get quotes from two bodies. These are planning estimates based on comparable ISO projects, not published benchmarks.
Common mistakes
- A separate AI binder. Running 42001 as a standalone system doubles your audits and reviews. Integrate it.
- Scoping too wide. "All AI in the group" on day one guarantees a long project. Start with one site or one function.
- Ignoring embedded and SaaS AI. If it influences your product or people, it is in scope, whoever built it.
- Impact assessment as a form exercise. It must consider the people affected, including operators, maintenance staff and customers, not just business risk.
- Assuming certification equals AI Act compliance. It does not. It helps you organize the evidence. Map the gaps with the framework crosswalk.
Monday-morning checklist
- Ask your current ISO registrar whether they are accredited for ISO/IEC 42001 and what a combined audit would look like.
- Decide your role: AI user only, or also provider (do you sell anything with AI in it?).
- Draft a one-paragraph scope: which site, which functions, which AI systems.
- Add "AI system" as a change type in MOC and a nonconformance type in CAPA.
- Add AI to the next management review agenda.
- Start the AI inventory and use it to build a draft Statement of Applicability.
Note: This page summarizes the standard at a high level. Buy the standard from ISO, IEC or your national standards body before implementing, and confirm any regulatory use with counsel.
Sources and further reading
- ISO/IEC 42001:2023 Artificial intelligence - Management system (IEC Webstore)
- ISO/IEC 42006:2025 Requirements for bodies providing audit and certification of AI management systems (ISO)
This guide is general information for manufacturers, not legal advice. Confirm regulatory obligations with qualified counsel.
FAQ
Frequently asked questions
Is ISO/IEC 42001 certification required by the EU AI Act?
No. The EU AI Act does not require ISO/IEC 42001 certification. High-risk AI providers need a quality management system, and a 42001-based system can carry much of that evidence, but the Act relies on its own harmonized standards for presumption of conformity. Treat 42001 as a good organizational backbone, not a legal compliance certificate.
How long does ISO 42001 certification take for a mid-sized manufacturer?
If you already hold ISO 9001 or ISO/IEC 27001 and keep the scope narrow, plan on roughly six to twelve months from kickoff to certification audit. Most of the time goes into building the AI inventory, running impact assessments and producing a few months of operating records, such as reviews and internal audits, that auditors expect to see.
Can ISO 42001 be audited together with ISO 27001 or ISO 9001?
Yes, in most cases. Because the standards share the same high-level structure, many certification bodies offer integrated or combined audits. You keep one management review, one internal audit program and one document control process, and add the AI-specific clauses and Annex A controls. Confirm with your certification body that it is accredited for 42001.
Does ISO 42001 apply if we only buy AI and do not build it?
Yes. The standard applies to organizations that provide or use AI systems. A manufacturer that only deploys vendor AI, such as vision inspection or a GenAI copilot, can scope its management system around use, supplier controls and impact assessment, and leave out development controls that do not apply, with a justification.
Related guides
Frameworks
NIST AI RMF for manufacturers: Govern, Map, Measure, Manage on the plant floor
The NIST AI Risk Management Framework is a free, voluntary playbook for managing AI risk.
Guide · 7 min read
Frameworks
EU AI Act for manufacturers: roles, risk tiers and the 2026 timeline
Most manufacturers are deployers under the EU AI Act, with a few hard prohibitions and AI literacy duties already in force.
Guide · 8 min read
Frameworks
ISO 27001 vs IEC 62443: which one a manufacturer needs, and how to run both
ISO/IEC 27001 is a certifiable management system standard for information security across the whole company.
Guide · 7 min read
Frameworks
AI and OT security framework crosswalk for manufacturers
No single framework covers AI governance and OT security.
Guide · 8 min read