Skip to content
Blentera

Resources, Standards

ISO/IEC 42001 and the NIST AI RMF: a practical comparison

ISO/IEC 42001 and the NIST AI Risk Management Framework are the two most cited frameworks for governing AI. This guide explains what each one is, how they relate to the EU AI Act and its new harmonised standards, and how to choose.

Updated 20 November 2026. 9 minute read.

In short

  • ISO/IEC 42001:2023 is a certifiable management system standard for AI. It follows the same structure as ISO/IEC 27001, and its Annex A lists reference controls that a company selects and justifies.
  • The NIST AI RMF 1.0 (January 2023) is a free, voluntary framework built on four functions: Govern, Map, Measure and Manage. Its Generative AI Profile (NIST AI 600-1, July 2024) covers twelve risks specific to or made worse by generative AI.
  • ISO/IEC 23894:2023 is guidance on AI risk management based on ISO 31000. It is not certifiable.
  • None of the three gives a presumption of conformity with the EU AI Act. That role belongs to harmonised standards from CEN-CENELEC JTC 21. The first, EN 18286:2026 on quality management, was cited in the Official Journal in October 2026.

ISO/IEC 42001: the AI management system standard

ISO/IEC 42001:2023 specifies requirements for establishing, implementing, maintaining and continually improving an artificial intelligence management system (AIMS). ISO and IEC published it in December 2023 through the joint committee ISO/IEC JTC 1/SC 42. It applies to any organisation that provides or uses AI-based products or services, whatever its size or sector.

It uses the harmonised structure shared by ISO management system standards, such as ISO/IEC 27001 for information security and ISO 9001 for quality. Clauses 4 to 10 cover context, leadership, planning, support, operation, performance evaluation and improvement. A company that already runs an ISO/IEC 27001 system will recognise most of the mechanics: policy, risk assessment, internal audit and management review.

Annex A controls in general terms

Annex A is a normative list of reference controls. The 2023 edition has 38 controls grouped under nine objectives:

  • policies related to AI;
  • internal organisation, such as roles and responsibilities;
  • resources for AI systems, including data, tools, computing and human resources;
  • assessing the impacts of AI systems on individuals, groups and society;
  • the AI system life cycle, from design to operation and monitoring;
  • data for AI systems, such as quality and provenance;
  • information for interested parties;
  • use of AI systems, including intended use and responsible use;
  • third-party and customer relationships.

A company does not have to apply every control. It selects controls through its risk treatment, and records which apply and why others are excluded in a statement of applicability. Annex B gives implementation guidance for each control.

Certification under ISO/IEC 42001

ISO does not certify companies. Independent certification bodies audit an AIMS against ISO/IEC 42001 and issue certificates. Certification is voluntary.

ISO/IEC 42006:2025, published in July 2025, sets out additional requirements for bodies that audit and certify AI management systems. It builds on ISO/IEC 17021-1, the general standard for management system certification bodies. When choosing an auditor, check that it is accredited for ISO/IEC 42001 by a national accreditation body.

  • Define the scope: which AI systems, roles and sites the AIMS covers.
  • Build the system: AI policy, risk and impact assessments, statement of applicability and records.
  • Run it for a period, with an internal audit and a management review.
  • Pass the external certification audit, then the regular surveillance audits.

A certificate shows that a management system meets ISO/IEC 42001 within its scope. It does not show that a specific AI system is safe or that the company meets the EU AI Act.

The NIST AI RMF and its Generative AI Profile

The US National Institute of Standards and Technology released the AI Risk Management Framework (AI RMF 1.0) on 26 January 2023. It is voluntary, free and not tied to any law, and it is designed for organisations of any size or sector.

Part 1 describes the characteristics of trustworthy AI: valid and reliable; safe; secure and resilient; accountable and transparent; explainable and interpretable; privacy-enhanced; and fair, with harmful bias managed. Part 2, the Core, organises the work into four functions:

  • Govern: a culture of risk management, with policies, roles, accountability and oversight. It applies across the other three.
  • Map: establish the context of each AI system, its purpose, users and potential impacts.
  • Measure: assess, analyse and track the identified risks with suitable methods and metrics.
  • Manage: prioritise and act on risks, respond to incidents and communicate.

NIST also publishes a Playbook with suggested actions for each subcategory.

The Generative AI Profile

NIST AI 600-1, published on 26 July 2024, is a companion profile for generative AI. It describes twelve risks that are unique to or made worse by generative AI: CBRN information or capabilities; confabulation; dangerous, violent or hateful content; data privacy; environmental impacts; harmful bias and homogenisation; human-AI configuration; information integrity; information security; intellectual property; obscene, degrading or abusive content; and value chain and component integration. It maps more than 200 suggested actions to the four functions.

NIST states on its AI Resource Center that a revision of the AI RMF is in progress. As of November 2026, AI RMF 1.0 remains the published version. Check NIST's site for updates before you build on specific wording.

ISO/IEC 23894: guidance on AI risk management

ISO/IEC 23894:2023, published in February 2023, gives guidance on how organisations that develop, deploy or use AI can manage AI-specific risks. It adapts the principles, framework and process of ISO 31000, the general risk management standard, to AI.

It is a guidance document, so it contains recommendations rather than requirements, and a company cannot be certified against it. It is most useful as the method behind the risk assessment that ISO/IEC 42001 requires. Its annexes map AI risk management activities to the AI system life cycle and list common AI risk sources and objectives.

Two related standards complete the picture. ISO/IEC 42005:2025 gives guidance on AI system impact assessments, and ISO/IEC 42006:2025 covers certification bodies.

How they relate to the EU AI Act and harmonised standards

Under Article 40 of the AI Act, high-risk AI systems that conform to harmonised standards whose references are published in the Official Journal are presumed to meet the requirements those standards cover. ISO/IEC 42001, ISO/IEC 23894 and the NIST AI RMF are not harmonised standards. Following them can support AI Act work, but it gives no presumption of conformity.

CEN-CENELEC JTC 21

The Commission asked the European standardisation organisations CEN and CENELEC to develop harmonised standards for the AI Act. The work sits with their Joint Technical Committee 21 on artificial intelligence. Status as of November 2026:

  • EN 18286:2026, Quality management system for EU AI Act regulatory purposes, became available in July 2026. It is the first JTC 21 standard finished for the AI Act.
  • In October 2026, the Commission cited EN 18286:2026 in the Official Journal through Implementing Decision (EU) 2026/2274. According to CEN-CENELEC and the Commission, providers of high-risk AI systems that follow it are presumed to meet the quality management system requirements of Article 17.
  • Other standards are still in development, including drafts on AI risk management (prEN 18228), cybersecurity for AI (prEN 18282) and logging (prEN 18229-1). Their timing is not settled.

CEN-CENELEC describes EN 18286 as designed to sit on top of existing frameworks such as ISO/IEC 42001 and ISO 9001 rather than replace them. A company with an ISO/IEC 42001 system therefore has a head start, but must still close the gaps to the AI Act.

Most harmonised standards serve providers of high-risk AI systems. Under the Digital Omnibus, Regulation (EU) 2026/1744, high-risk obligations apply from 2 December 2027 for Annex III use cases and 2 August 2028 for Annex I products. Companies that only deploy agents for tasks that are not high-risk will use these standards mainly as good practice.

How to choose between them

The frameworks are not rivals. They answer different questions, and many companies use more than one.

  • Choose ISO/IEC 42001 if customers or partners ask for an independent certificate, or if you already run ISO/IEC 27001 or ISO 9001 and want AI inside the same system.
  • Choose the NIST AI RMF if you want a free, flexible starting point without certification, or if you sell to US customers who use it.
  • Use the Generative AI Profile alongside either one if you deploy large language models or agents.
  • Use ISO/IEC 23894 as the method for AI risk assessment, whichever framework you choose.
  • If you provide high-risk AI systems in the EU, follow the harmonised standards as they are cited, starting with EN 18286.

A pragmatic path for a company using agents

  1. Start with an inventory and a register of AI use cases, with owners and risk levels.
  2. Use the four NIST functions to structure the first year of work.
  3. Write an AI policy and run risk and impact assessments with ISO/IEC 23894 and ISO/IEC 42005 as guides.
  4. Decide on ISO/IEC 42001 certification once the system works and there is a business case.

Whichever framework you choose, the evidence is the same: who owns each AI system, what it may do, how risks were assessed, and what it actually did.

How Blentera helps

Blentera is a platform where companies run AI agents with employees in charge. Each agent has a role in the org chart, approval rules decide what needs an employee's approval, and each agent can use only the apps and tools it is given. Budgets warn at 80% and pause the agent at the limit. The EU AI Act use-case register records each use case, and the activity log, which can be exported, gives evidence of what agents did. It is hosted in the EU. These records support the evidence the frameworks above ask for. Using Blentera does not make a company certified to any standard or compliant with the AI Act.

Questions

Sources

  1. ISO/IEC 42001:2023 AI management systems, ISO
  2. ISO/IEC 42001 explained, ISO
  3. ISO/IEC 42006:2025 Requirements for bodies providing audit and certification of AI management systems, ISO
  4. ISO/IEC 23894:2023 Guidance on risk management, ISO
  5. ISO/IEC 42005:2025 AI system impact assessment, ISO
  6. AI Risk Management Framework, NIST
  7. Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1
  8. Generative Artificial Intelligence Profile (NIST AI 600-1), NIST
  9. AI RMF resources and revision status, NIST AI Resource Center
  10. EN 18286 in the spotlight: supporting compliance with the AI Act, CEN-CENELEC, 30 July 2026
  11. First ever European standard on AI published in the OJEU, CEN-CENELEC
  12. First harmonised standard supporting the AI Act published in the Official Journal, European Commission
  13. Standardisation of the AI Act, European Commission
  14. Regulation (EU) 2024/1689 (AI Act), consolidated text of 27 July 2026, EUR-Lex

This guide is general information, not legal advice. Check how the rules apply to your company with your own advisers.

Get started

Put your first agent to work in minutes.

Three steps, and your first agent is working for your company.

  1. Step 1

    Create an account

    Sign up with your work email and invite your team. No sales call needed.

  2. Step 2

    Add your first agent

    Create an agent, give it a role, and connect your own AI key and the apps it may use.

  3. Step 3

    Give it a task

    Ask in Blentera or your chat app, and follow the work until it is done.