
ISO 420017 min read
ISO 42001 vs EU AI Act: what differs and what overlaps
ISO 42001 is a voluntary management standard, the EU AI Act is binding law. See where they differ, where they overlap and what a certificate covers.
By Michael Schwarz, Lead auditor ISO 9001 and 14001, SICE
ISO 42001 is a voluntary management system standard for organisations that develop or use AI. The EU AI Act is binding law that regulates AI systems by risk class, and an ISO 42001 certificate does not replace compliance with it.
The two documents answer different questions. The standard asks how you manage AI in your organisation. The law asks what your specific AI system may do and what you must prove about it. A business that handles the first well has a head start on the second, but the work does not end there.
The short answer
| ISO 42001:2023 | EU AI Act | |
|---|---|---|
| Nature | Voluntary international standard | Regulation (EU) 2024/1689, binding law |
| Object | The management system of an organisation | AI systems and general-purpose AI models |
| Approach | Process: policy, risk, controls, audit, improvement | Risk classes: prohibited, high-risk, transparency duties, minimal risk |
| Proof | Certificate, internal audit, management review | Conformity assessment, technical documentation, registration, market surveillance |
| Who asks for it | Customers, tenders, partners | Authorities |
If a customer's supplier questionnaire asks for an AI governance framework, ISO 42001 is the usual answer. If a regulator asks whether your recruitment tool may be placed on the EU market, only the Act answers that.
What ISO 42001 requires
ISO 42001 specifies an AI management system. The official description is on the ISO page for the standard. It follows the high level structure used by other management system standards: context of the organization (clause 4), leadership (5), planning (6), support (7), operation (8), performance evaluation (9) and improvement (10).
In practice that means you:
- define the scope: which AI systems you build, buy or use, and in which roles
- set an AI policy and assign responsibility at management level
- assess AI risks and the impact of AI systems on people, then decide how to treat them
- select controls from the annexes of the standard and document why you chose them
- train staff, keep documented information and manage suppliers of AI components
- audit yourself internally, review the results in management and correct what failed
The standard does not tell you which AI systems are allowed. It tells you how to run a system of control around them. That is why it fits a five-person agency using AI tools as well as a software company training its own models.
What the EU AI Act requires
The Act sorts AI systems into risk classes. Certain practices are prohibited outright. High-risk systems, for example in recruitment, credit scoring, education or critical infrastructure, carry the heaviest duties. Systems that interact with people or generate content carry transparency duties. Everything else is largely left alone. Separate rules apply to providers of general-purpose AI models.
The obligations depend on your role:
- Provider: you develop a system or have it developed and place it on the market under your name. For high-risk systems this brings a risk management system, data governance, technical documentation, logging, human oversight, a quality management system and a conformity assessment.
- Deployer: you use a system in your own business. For high-risk systems you must use it according to the instructions, assign human oversight, monitor operation, keep logs and inform affected people and employees where required.
- Importer and distributor: you bring systems from outside the EU or pass them on, and you check that the provider has done their part.
Some duties apply to everyone. The ban on prohibited practices and the duty to ensure sufficient AI literacy among staff have applied since 2 February 2025. Rules for general-purpose AI models followed on 2 August 2025. Most remaining provisions, including the high-risk requirements, are scheduled from 2 August 2026, with product-related high-risk rules a year later. Postponements have been proposed, so check the current timetable before you plan around a date.
Where the two overlap
The Act asks for several things that a working AI management system produces anyway:
- Risk management: the Act requires a continuous risk management process for high-risk systems. ISO 42001 builds risk assessment and treatment into planning and operation (clauses 6 and 8).
- Quality management: providers of high-risk systems need a documented quality management system. A certified management system is a ready structure for it.
- Documentation and records: technical documentation and logs under the Act map onto the documented information the standard requires in clause 7.
- AI literacy: the duty to train staff corresponds to competence and awareness requirements in the same clause.
- Oversight and accountability: named owners, management review and internal audit make human oversight and responsibility traceable.
This overlap is the reason many companies start with the standard. You build the structure once and attach the legal requirements to it afterwards.
Where they differ
Three differences matter in practice.
Object. The standard certifies an organisation's management system. The Act regulates individual systems. A company can have a clean management system and still ship a system that falls into a prohibited category.
Depth. The Act contains specific technical and procedural requirements for high-risk systems, such as data quality criteria, accuracy and robustness, and a formal conformity assessment. ISO 42001 stays at management level and leaves the technical detail to you.
Consequence. Non-compliance with the Act can lead to action by market surveillance authorities and to fines. Not holding an ISO 42001 certificate has no legal consequence. It may cost you a tender or a customer audit.
Does ISO 42001 prove compliance with the AI Act?
No. ISO 42001 is not a harmonised standard under the AI Act. Harmonised standards are the ones that give a presumption of conformity once they are cited in the Official Journal of the EU, and ISO 42001 does not have that status. A certificate shows that you run a structured AI management system. It does not show that a particular system meets the Act.
For a customer or a purchasing department, that is still a meaningful signal. It tells them you have a policy, a risk process and an audit cycle for AI. Just do not describe it as legal compliance in your offers.
Does the AI Act apply to a Swiss company?
It can. The Act applies to providers that place AI systems on the EU market, and to providers and deployers outside the EU when the output of their system is used in the EU. A Swiss software firm selling an AI feature to a German customer is in scope. A Swiss agency using a chatbot only for Swiss clients is in a different position, although its EU customers may still push contract terms down to it.
For companies in Germany, Austria and France the Act applies directly.
Checklist: what to do first
This order works for most small and medium businesses.
How Zertify fits in
If customers ask for proof of AI governance, a certificate against the standard answers that request. The ISO 42001 page describes the scope and what the assessment covers. You complete an online self-assessment against the requirements. A member of the SICE Certification Board reviews every assessment before the certificate is issued by Zertify under licence from SICE. The certificate carries a QR code that leads to a public verification page, so a buyer can check it without contacting you.
There is no on-site audit. If you want to understand how that differs from the classic route with an on-site audit by a certification body, the guide on ISO certification without an audit explains the trade-offs. For smaller teams, the overview of ISO certification for small and medium businesses covers how much documentation is realistic.
The certificate is issued within 4 hours after a passed assessment and payment. It is valid for three years and confirmed every year with a short online re-audit, which is a yearly subscription. Current prices are on the pricing page. When you are ready, you can start the assessment.
A Zertify certificate documents your AI management system. It does not replace the conformity assessment the AI Act requires for high-risk systems, and it should not be presented that way.
Frequently asked questions
Is ISO 42001 mandatory under the EU AI Act?
No. The AI Act does not require ISO 42001 certification. The standard is voluntary. It can support your compliance work, for example for risk management and documentation, but it is not a legal requirement.
Does an ISO 42001 certificate make me compliant with the EU AI Act?
No. ISO 42001 is not a harmonised standard under the AI Act, so it does not give a presumption of conformity. The certificate shows that you run a structured AI management system. Whether a specific AI system meets the Act depends on its risk class and the duties that follow from it.
Does the EU AI Act apply to companies in Switzerland?
It can. Swiss providers and deployers fall under the Act if they place AI systems on the EU market or if the output of their system is used in the EU. A Swiss company serving only Swiss customers is not directly in scope, but EU customers may pass requirements on by contract.
Which should a small business tackle first, the standard or the law?
Start with an inventory of your AI systems and a check against the Act's prohibited and high-risk categories. That shows what is legally required of you. Then use ISO 42001 as the structure for policy, risk assessment and records, especially if customers ask for proof.
Can I use ISO 42001 for the AI literacy duty?
It helps. The standard requires competence and awareness of the people who work within your AI management system, and training records support the AI literacy duty in the Act. The Act sets its own expectation that staff have sufficient knowledge for the AI systems they use.


