MAP
5 categoriesKnow the context and the risks- 1Context
- 2Categorization
- 3Capabilities and goals
- 4Component risks
- 5Impacts
Context · Categorization · Capabilities and goals · Component risks · ImpactsMEASURE
4 categoriesAssess, analyze and track them- 1Methods and metrics
- 2Trustworthiness
- 3Risk tracking
- 4Measurement feedback
Methods and metrics · Trustworthiness · Risk tracking · Measurement feedbackMANAGE
4 categoriesPrioritize and act- 1Prioritize and respond
- 2Benefits and impacts
- 3Third-party risks
- 4Response and recovery
Prioritize and respond · Benefits and impacts · Third-party risks · Response and recovery
GOVERN
Applies across all three6 categories- 1Policies and processes
- 2Accountability
- 3Workforce diversity
- 4Risk culture
- 5Engaging AI actors
- 6Supply chain
You choose to adopt it, often to find gaps in how you manage AI risk.
Anyone who designs, deploys or uses AI systems. It’s voluntary guidance, not law.
GOVERN 6 covers AI risks from third-party software, data and supply chains.
Anyone running AI built by others, including AI features inside software you already use.
NIST AI 600-1 adds actions for 12 generative AI risks, mapped to the RMF.
Anyone developing or using generative AI, from chat assistants to content and code tools.
It counts substantial compliance with NIST’s generative AI profile toward one defense.
Anyone doing business in Texas, where the state’s AI law took effect January 1, 2026.
GOVERN 1
Put policies, processes and practices in place for mapping, measuring and managing AI risk. GOVERN is cross-cutting: it runs through and enables the other three functions.
GOVERN 1.6
Have mechanisms to inventory AI systems, resourced according to your risk priorities. You can’t map, measure or manage AI you haven’t found.
GOVERN 2
Put accountability structures in place, so the right teams and people have the authority, the responsibility and the training for AI risk work.
GOVERN 6
Have policies and procedures for the AI risks and benefits that come from third-party software and data, and from other supply chain issues.
MAP 1 · MAP
Record what the system is for, where and how it will be used, and the laws, norms and expectations that apply (MAP 1.1). The same model can carry different risks in different settings.
MAP 5 · MAP
Characterize how the system could affect individuals, groups, communities, organizations and society. Those impacts shape what you measure next.
MEASURE 2
Evaluate systems for trustworthy characteristics, with the methods and metrics chosen under MEASURE 1. Document the test sets, metrics and tools used (MEASURE 2.1).
MEASURE 3
Have mechanisms to track identified risks over time, including unanticipated and emergent ones, and how the system performs once it is deployed.
MANAGE 1
Decide whether the system meets its intended purpose and should go ahead (MANAGE 1.1). Then prioritize risks using what MAP and MEASURE found, and respond to them.
Not as of October 2026. NIST says AI RMF 1.0 “is being revised as part of the White House AI Action Plan,” but it hasn’t published a revised text or set a date. [1]
Since January 1, 2026, Texas’s AI law shields a defendant who finds a violation through internal review while substantially complying with NIST’s generative AI profile, or another recognized framework. [5]
NIST calls the Playbook “neither a checklist nor set of steps.” It suggests voluntary actions for each subcategory, so you take the ones that fit your context. [3]
Is the NIST AI RMF mandatory?
No. NIST released it for voluntary use, and there’s no certification. It still turns up in law: Texas’s AI law, known as TRAIGA, treats substantial compliance with NIST’s generative AI profile as part of one defense against liability. Organizations also adopt it by choice, often to find gaps in how they manage AI risk and to give their teams one shared way to talk about it.
What are the four functions of the NIST AI RMF?
GOVERN sets the policies and accountability for managing AI risk, and it runs through the other three. MAP establishes each system’s context and its possible impacts. MEASURE tests and tracks the risks, with chosen methods and metrics. MANAGE prioritizes the risks, treats them and plans the response. Together they hold 19 categories: six in GOVERN, five in MAP, and four each in MEASURE and MANAGE.
What is the NIST Generative AI Profile (AI 600-1)?
It’s a companion to the AI RMF, published July 26, 2024, not a replacement. It names 12 risks unique to or made worse by generative AI: chemical, biological, radiological or nuclear (CBRN) information, confabulation, dangerous or hateful content, data privacy, environmental impacts, harmful bias, human-AI configuration, information integrity, information security, intellectual property, obscene or abusive content, and value chain integration. It then maps suggested actions to the framework’s subcategories.
When will NIST publish the revised framework?
NIST hasn’t said. America’s AI Action Plan of July 2025 asked it to revise the framework, and NIST says the Playbook will be updated once that’s done. Until then, version 1.0 and the generative AI profile are current. In April 2026, NIST also released a concept note for a critical infrastructure profile. In Enzai, frameworks have versions, so you choose when an assessment switches to a new one.
Can you get certified against the NIST AI RMF?
No. There’s no certification for the AI RMF. If you need a certificate, ISO/IEC 42001 is the certifiable standard for an AI management system, and accredited certification bodies issue the certificates. The two cover similar ground, including accountability, risk assessment and risk treatment, so you can run them as one program rather than doing the same work twice.
How does Enzai help with the NIST AI RMF?
You can build the AI RMF as your own framework in Enzai, with its subcategories as controls, then assess each AI system against it. Classify a system first, and only the controls that fit it are created. Owners, due dates and approvals support GOVERN. The inventory supports GOVERN 1.6 and MAP. Metrics recorded as evidence support MEASURE, and the risk register, with mitigations and proof, supports MANAGE.
Sources · Reviewed Oct 2026
