NIST AI RMF
NIST AI Risk Management Framework (AI RMF): functions, playbook and generative AI profile
In short
The NIST AI Risk Management Framework (AI RMF 1.0) is a voluntary US framework, released on 26 January 2023, that helps organisations identify, assess and manage risks from AI systems. It is organised around four functions, Govern, Map, Measure and Manage, and is supported by a Playbook and a Generative AI Profile (NIST AI 600-1).
Key takeaways
- The AI RMF is voluntary, sector-neutral and not certifiable. It gives you a shared structure and vocabulary, not a pass or fail test.
- Its Core has four functions: Govern, Map, Measure and Manage. Govern applies across the whole organisation; the other three apply to individual AI systems.
- NIST AI RMF 1.0 contains 72 subcategories across 19 categories. NIST states that the actions are not a checklist or an ordered set of steps.
- The Playbook (suggested actions and references) and NIST AI 600-1 (generative AI risks) are companions, not replacements for the framework.
- NIST has said AI RMF 1.0 is being revised as part of the White House AI Action Plan, so check the NIST page for the current version before building long-lived documentation around it.
What is the NIST AI RMF?
The NIST AI RMF is a risk management framework published by the US National Institute of Standards and Technology as NIST AI 100-1. Version 1.0 was released on 26 January 2023 after a public consultation with industry, academia and civil society.
It is voluntary. It does not create legal obligations on its own, and there is no official NIST certification. Organisations use it to structure how they think about AI risk and to show regulators, customers and auditors that they follow a recognised approach.
The framework is deliberately technology-neutral and sector-neutral. It applies to organisations that design, develop, deploy or use AI systems, at any size.
What are the four functions of the NIST AI RMF?
The four functions of the AI RMF Core are Govern, Map, Measure and Manage. Govern sets up the culture, policies and accountability that everything else depends on; Map, Measure and Manage are applied to specific AI systems and lifecycle stages.
| Function | Purpose | Typical outputs |
|---|---|---|
| Govern | Build the policies, roles, accountability and risk culture for AI across the organisation. Applies to all stages. | AI policy, named owners, risk tolerance, training, an AI system inventory process, decommissioning rules |
| Map | Establish context: what the system is for, who it affects, what could go wrong, and what its limits are. | Intended purpose, stakeholder and impact analysis, third-party and data dependencies, risk identification |
| Measure | Analyse and track risks with suitable metrics, tests and evaluation methods. | Test plans and results, bias and robustness measures, monitoring metrics, documented limitations |
| Manage | Prioritise and act on the mapped and measured risks, then monitor and respond over time. | Risk treatment decisions, incident response, change management, residual risk acceptance |
The framework describes the functions as iterative. Most users start with Govern, then Map, and move between Measure and Manage as the system and its context change.
How many categories and subcategories does the AI RMF have?
AI RMF 1.0 has 19 categories and 72 subcategories in total, counted from the tables in NIST AI 100-1: Govern has 6 categories and 19 subcategories, Map 5 and 18, Measure 4 and 22, and Manage 4 and 13.
Each subcategory is an outcome statement, for example GOVERN 1.6: mechanisms are in place to inventory AI systems, resourced according to organisational risk priorities. You choose which outcomes matter for your context; NIST says organisations may select from the categories and subcategories or apply all of them.
What are the characteristics of trustworthy AI in the AI RMF?
The framework describes trustworthy AI as having seven characteristics, which you balance against each other according to the context of use.
- Valid and reliable
- Safe
- Secure and resilient
- Accountable and transparent
- Explainable and interpretable
- Privacy-enhanced
- Fair, with harmful bias managed
Valid and reliable is described as the foundation; the other characteristics depend on it. Trade-offs, such as privacy against explainability, are expected and should be documented.
What is the NIST AI RMF Playbook?
The NIST AI RMF Playbook is a companion resource that offers suggested actions, references and related guidance for each Govern, Map, Measure and Manage subcategory. It is voluntary, is not a checklist, and you do not have to follow every suggestion.
The first full version appeared on 30 March 2023 and NIST has continued to update it; the NIST page shows an update on 10 June 2026. Use it to turn an outcome such as 'mechanisms are in place to inventory AI systems' into concrete tasks, owners and evidence.
What is the NIST Generative AI Profile (NIST AI 600-1)?
NIST AI 600-1, the Generative Artificial Intelligence Profile, was released on 26 July 2024. It is a cross-sectoral companion to AI RMF 1.0 that helps organisations identify risks specific to generative AI and suggests actions to manage them, mapped to the Core.
Use it alongside the main framework if you build on or deploy large language models, image generators or similar systems. It adds generative-specific considerations, such as confabulation, information integrity and harmful content, to the standard Govern, Map, Measure and Manage outcomes.
How do you implement the NIST AI RMF in practice?
Start by treating the framework as a structure for your own governance programme, then work through it in this order.
- Govern first: agree an AI policy, name accountable owners and set your risk tolerance. Create a process to keep an AI system inventory (GOVERN 1.6).
- Map each system: record purpose, users, affected people, data sources and third-party components, and the context in which it will run.
- Measure: choose tests and metrics for each trustworthiness characteristic that matters for the system, and record the results and known limitations.
- Manage: decide treatments, accept or reject residual risk with an authorised approver, and define incident and change-management routes.
- Keep evidence: link policies, test results and decisions to each system so you can show what you did and when.
- Review on a schedule and when the system, data or context changes. Revisit the framework text itself, given the planned revision.
How does the NIST AI RMF relate to ISO/IEC 42001 and the EU AI Act?
The AI RMF is a voluntary risk framework, ISO/IEC 42001 is a certifiable management system standard, and the EU AI Act is binding law. They complement each other, but meeting one does not mean you meet another.
Many teams use the AI RMF's vocabulary to shape risk work, ISO/IEC 42001 to run it as a management system, and the AI Act to decide which systems carry legal duties.
| Aspect | NIST AI RMF | ISO/IEC 42001 | EU AI Act |
|---|---|---|---|
| Type | Voluntary framework | Management system standard | Regulation |
| Certifiable | No | Yes, via certification bodies | Conformity assessment for high-risk systems |
| Focus | Risk outcomes across the AI lifecycle | Organisation-wide AI management system | Obligations by risk category and role |
| Binding | No | No, unless contracted or required by a customer | Yes, in the EU |
How AIGovernr helps
AIGovernr includes a versioned NIST AI RMF framework release that maps published obligations to controls and evidence in a project workspace, alongside an AI system register, governance plans and evidence records. Coverage figures measure how much published content is mapped, not your own controls or evidence, and they are not certification or legal validation.
Frequently asked questions
Is the NIST AI RMF mandatory?
No. The NIST AI RMF is voluntary and does not itself create legal obligations. Customers, contracts and some US laws may nonetheless refer to it, so it can matter in practice. Check the specific law or contract that applies to you.
Can you get certified against the NIST AI RMF?
There is no official NIST AI RMF certification. Any certificate you see is issued by a third party under its own scheme. If you need a certifiable standard, ISO/IEC 42001 is the AI management system standard with accredited certification, and many organisations map the two together.
When was NIST AI RMF 1.0 released, and is it being updated?
AI RMF 1.0 was released on 26 January 2023. NIST states on its AI RMF page that the framework is being revised as part of the White House AI Action Plan. Confirm the current version on the NIST site before you finalise policies or mappings that cite specific subcategory numbers.
What is the difference between the AI RMF and the Playbook?
The AI RMF defines the outcomes: functions, categories and subcategories. The Playbook is a companion that suggests actions, references and further reading for each subcategory. The Playbook is optional guidance and not a checklist; you choose the suggestions that fit your system and risk level.
Does the AI RMF apply to generative AI?
Yes. The framework is technology-neutral, and NIST published the Generative AI Profile, NIST AI 600-1, on 26 July 2024 to cover risks particular to generative AI. Apply the profile alongside the four functions when you build or deploy language models or other generative systems.
Where should a small team start with the AI RMF?
Begin with Govern: a short AI policy, a named owner, and an inventory of the AI systems you use or build. Then map your highest-risk system, choose a few tests, and record decisions. NIST says you can select categories and subcategories rather than apply all 72 at once.
How does the NIST AI RMF help with the EU AI Act?
It does not make you compliant, but its structure supports similar work: inventorying systems, assessing risk, testing, human oversight and monitoring. Treat it as a way to organise evidence, then check each EU AI Act article that applies to your systems and role separately.
Sources
Facts on this page were checked against these primary sources on 3 October 2026. This guide is general information, not legal advice.
Put this into practice
Register your AI systems, see which obligations apply, and keep controls and evidence in one workspace.
Start free