What is ISO/IEC 42001 and who is it for?
ISO/IEC 42001:2023 specifies the requirements for an artificial intelligence management system - the rules, roles and processes with which an organization creates, deploys and uses AI systems responsibly. It was published in December 2023 and follows the harmonized high-level structure shared by ISO 9001, ISO/IEC 27001 and the other management system standards, so it integrates with them.
The standard is for any organization that has a role in relation to AI: a developer or provider of models and applications, a company that deploys AI in its own processes or products, a data provider, or an organization that uses third-party AI systems for decisions affecting people. The first step under the standard is precisely to define your role - the risks, obligations and controls depend on it.
- Developers and providersCompanies that create AI models, applications or services for customers - including as part of another product.
- Organizations that deploy AIUse AI for recruitment, credit decisions, customer service, medical or technical analysis.
- Suppliers to large customersContracts with corporate and public-sector customers increasingly ask for evidence of AI risk management.
- Exporting companiesPartners outside Bulgaria expect a familiar, international language of accountability and control.
What does the standard require?
As with the other management system standards, top management sets a policy and objectives, the organization assesses risks and plans controls, provides resources and competence, operates and measures, audits internally and improves. What is specific to ISO/IEC 42001 is the subject: AI systems and their effects on people, groups and society.
- AI risk assessment - for every system: what harm it may cause, how likely, to whom; and risk treatment with specific controls.
- Impact assessment - a documented analysis of the consequences of the system for individuals and society, including fairness, safety, privacy and transparency.
- Life cycle - requirements and objectives in design, verification and validation, deployment, monitoring in operation and retirement.
- Data management - provenance, quality, preparation and protection of the data used to train and operate the systems.
- Information for interested parties - what is communicated to users, customers and regulators, including about incidents.
- Responsible use and third parties - rules of use, human oversight and responsibilities with suppliers and customers.
The controls in Annex A of the standard are selected according to the risk assessment and documented in a statement of applicability - the same model as ISO/IEC 27001. That is why organizations that already hold an information security certificate implement ISO/IEC 42001 considerably faster.
AI Act: which obligations and from when?
Regulation (EU) 2024/1689 on artificial intelligence has been in force since 1 August 2024 and applies in stages. In 2026 Regulation (EU) 2026/1744 (the “Digital Omnibus on AI”) postponed the deadlines for high-risk systems without changing the obligations themselves.
- 2 February 2025
The prohibited practices (Article 5) and the AI literacy obligation (Article 4) - providers and deployers ensure a sufficient level of knowledge among the staff working with AI.
- 2 August 2025
The obligations for providers of general-purpose AI models.
- 2 August 2026
General application of the Regulation, including the transparency obligations under Article 50 - chatbots, generated content, deepfakes.
- 2 December 2027
High-risk systems under Annex III - biometrics, critical infrastructure, education, employment, access to essential services, law enforcement, migration, justice.
- 2 August 2028
High-risk systems embedded in products under Annex I - machinery, medical devices, lifts, toys, vehicles and other regulated products.
For high-risk systems, providers must have a risk management system (Article 9), data governance (Article 10), technical documentation and record-keeping (Articles 11 and 12), human oversight (Article 14) and a quality management system (Article 17). Deployers use them in accordance with the instructions, ensure oversight by competent people and monitor their operation (Article 26). Serious incidents are reported to the market surveillance authorities (Article 73).
How does ISO/IEC 42001 help with the AI Act?
| AI Act | ISO/IEC 42001 |
|---|---|
| Provider’s quality management system (Article 17) | The whole management system - policy, roles, processes, documented information, internal audit, management review |
| Risk management throughout the life cycle (Article 9) | AI risk assessment and treatment, impact assessment, monitoring in operation |
| Data governance (Article 10) | The data controls - provenance, quality, preparation, protection |
| Technical documentation and records (Articles 11, 12) | Documented information for every AI system and its life cycle |
| Human oversight (Article 14) and deployer obligations (Article 26) | Rules for responsible use, roles and competence, oversight |
| AI literacy (Article 4) | Competence and awareness of staff (clauses 7.2 and 7.3) |
| Reporting of serious incidents (Article 73) | Incident process and information for interested parties |
An important clarification: the conformity of high-risk systems with the Regulation is demonstrated under its own rules - conformity assessment, CE marking and, for some systems, a notified body. An ISO/IEC 42001 certificate does not replace that procedure. It proves that the organization has a working system which produces the required documents, assessments and controls - the foundation on which conformity is built and maintained.
Do you need an ISO/IEC 42001 certificate?
The certificate makes sense when
- You develop or supply AI products and services to customers.
- You use AI for decisions about people - recruitment, credit, pricing, medicine, security.
- Customers, contracting authorities or partners ask for evidence of AI risk management.
- You fall within the scope of high-risk systems and want a system ready for 2027-2028.
Not yet necessary when
- You use AI only in office tools, without decisions affecting people.
- You have no AI systems of your own and your suppliers bear the responsibility for them.
- Even so, AI literacy under Article 4 and transparency under Article 50 apply to you too.
Describe every AI system you use or offer, your role in relation to it and whether it falls under Annex I or III of the Regulation.
For systems with an effect on people, carry out a risk assessment and an impact assessment - this is the core of both the standard and the Regulation.
How does ISO/IEC 42001 certification work?
As with the other standards: a two-stage audit - a review of the documentation and readiness, then verification of implementation on site; a three-year certificate with surveillance audits in the second and third year. Organizations certified to ISO/IEC 27001 can combine the two audits.
ISO/IEC 42001 is outside the scope of accreditation of CERTIND S.A. (RENAR SM 041). The certificate is issued without an accreditation mark - tell us if your customer requires an accredited certificate, so that you can judge in advance whether it serves your purpose.
More about the standard - on the ISO/IEC 42001 page; for the link between information security and the new European rules - in the article NIS 2 in Bulgaria and ISO/IEC 27001.
Questions about ISO/IEC 42001 and the AI Act
Is ISO/IEC 42001 mandatory under the AI Act?
No. The Regulation does not require certification to a standard. ISO/IEC 42001 is a voluntary standard whose management system covers a large part of the obligations of providers and deployers - risk management, data, documentation, human oversight - and serves as evidence for customers and partners.
From when do the obligations for high-risk AI systems apply?
After Regulation (EU) 2026/1744 the obligations for high-risk systems under Annex III apply from 2 December 2027, and for systems embedded in regulated products under Annex I - from 2 August 2028. The general provisions of the Regulation apply from 2 August 2026.
Can a company that only uses AI be certified to ISO/IEC 42001?
Yes. The standard applies to any role - developer, provider or an organization that uses AI systems. The scope of the management system is defined according to the role and the systems the organization uses.
Is the ISO/IEC 42001 certificate obtained through ESCERT accredited?
No. ISO/IEC 42001 is outside the scope of accreditation of CERTIND S.A. (RENAR SM 041) and the certificate is issued without an accreditation mark. The accredited standards are listed on the accreditation page.
Can ISO/IEC 42001 be integrated with ISO/IEC 27001?
Yes. The two standards share the same structure, and the controls are selected by risk assessment and documented in a statement of applicability. An organization certified to ISO/IEC 27001 can extend its system and combine the audits.