Frequently asked questions about the NIST AI RMF

From each FAQ, you can go to the page section with the full rule.

Disclaimer

This page gives general information. It is not legal advice.

What is the NIST AI Risk Management Framework (NIST AI RMF)?

The NIST AI RMF (NIST AI 100-1, January 2023) is voluntary guidance to help organizations manage AI risks. It is for organizations "designing, developing, deploying, or using AI systems" and the "trustworthy and responsible" development and use of AI systems. Its Core has four functions with categories and subcategories, and it gives seven characteristics of trustworthy AI.

As of October 2026, NIST reports that AI RMF 1.0 "is being revised as part of the White House AI Action Plan", but no new version is available. It sets no obligation. See The framework and its status.

What are the four core functions of the NIST AI RMF?

The four functions are GOVERN, MAP, MEASURE, and MANAGE, with 72 subcategories in NIST AI 100-1, Tables 1 to 4. GOVERN "applies to all stages of organizations' AI risk management processes and procedures". MAP, MEASURE, and MANAGE "can be applied in AI system-specific contexts and at specific stages of the AI lifecycle". After GOVERN, most users start with MAP.

If there is a governance structure, the functions can occur in a different sequence. NIST recommends that "the process should be iterative", and its actions are not "a checklist".

The Playbook gives different, voluntary suggested actions. For each function and the sequence conditions in Section 5, see The four functions of the Core.

Is the NIST AI RMF mandatory?

No. NIST AI RMF 1.0 is voluntary guidance, and GOVERN is part of it. It sets no obligation for organizations or AI actors. Laws such as the EU AI Act set obligations independently.

Texas H.B. 149 is in effect from January 1, 2026. Section 552.105(e)(2)(D) names AI 600-1, not AI RMF 1.0, in a defense to an attorney general action. The defendant finds a Chapter 552 violation through "an internal review process" while it "substantially complies with the most recent version" of AI 600-1. The provision also accepts the "most recent version" of a different "nationally or internationally recognized risk management framework for artificial intelligence systems". Compliance is not sufficient without that "internal review process", and there is no obligation to use NIST guidance. For the defense, its alternatives, and the action for a civil penalty or an injunction, see Legal effect.

Is the NIST AI RMF being updated?

Yes. As of October 2026, NIST reports: "The AI RMF 1.0 is being revised as part of the White House AI Action Plan". No new version is available. AI RMF 1.0 of January 2023 continues to be the version in use.

The Action Plan is policy, not law. It recommends removal of "references to misinformation, Diversity, Equity, and Inclusion, and climate change".

The Playbook revision follows the AI RMF revision. AI 100-1 also gives "no later than 2028" for "a review with formal input from the AI community". For these schedules and the 2022 drafts, see The framework and its status.

What is the NIST AI RMF Playbook?

The NIST AI RMF Playbook is a voluntary companion resource with suggested actions for the Core outcomes. The Core text and subcategories are in NIST AI 100-1, Tables 1 to 4. Organizations can select the suggested actions that apply to them: the Playbook "is neither a checklist nor set of steps to be followed in its entirety".

As of October 2026, NIST gives this schedule: "The Playbook will be updated after the AI RMF is revised". See The framework and its status.

How is the NIST AI RMF different from ISO/IEC 42001?

The NIST AI RMF gives voluntary AI risk management outcomes, while ISO/IEC 42001:2023 sets requirements for an AI management system. The crosswalk that NIST shows, with Microsoft as provider, maps Core subcategories to clauses and Annex B of the draft ISO/IEC FDIS 42001. The NIST entry "does not imply NIST endorsement" or that each instrument "comprehensively covers the contents of the other". For certification of an AIMS, see ISO/IEC 42001 scope and certification.

Modulos has different Frameworks for the instruments. See How NIST AI RMF compares to other frameworks.

What is the NIST Generative AI Profile (AI 600-1)?

NIST AI 600-1 (July 2024) is a voluntary cross-sectoral profile of AI RMF 1.0 for generative AI (GAI). It gives 12 GAI risks and suggested actions for only some Core subcategories. It "was developed per Section 4.1(a)(i)(A) of EO 14110", which Executive Order 14148 revoked on January 20, 2025.

Texas H.B. 149, Section 552.105(e)(2)(D), in effect from January 1, 2026, has defense conditions for a defendant in an attorney general action:

• A Chapter 552 violation found through "an internal review process".
• At that time, the defendant "substantially complies with the most recent version" of the Profile or a different "nationally or internationally recognized risk management framework for artificial intelligence systems".

These conditions give no obligation to use the Profile, and compliance is not sufficient without the "internal review process". Modulos has no Framework for AI 600-1. See the Profile's structure.

How do you operationalize the NIST AI RMF in practice?

Modulos recommends OFF-3 for the organization AI System and MFF-3 for each AI System, then GOVERN, MAP, MEASURE, and MANAGE in a cycle. NIST gives outcomes, not "a checklist", and lets the functions occur in a different sequence if there is a governance structure.

These conditions change that sequence or the work:

• No organization AI System: contact Modulos support before you add OFF-3. See AI System structure.
• Requirements in all four functions, with Govern 1.1, Govern 2.3, Map 5.1, and Measure 2.1 only in OFF-3: see The Requirements.
• The risk register does not record the risk response or if you accept a residual risk. Record the documentation with the Manage 1.3 and Manage 1.4 Requirements.
• Profiles and AI 600-1: work without the Frameworks.
• Changes to context, risks, or impacts: do the recommended sequence again.

What is the NIST AI RMF Govern function?

GOVERN is the NIST AI RMF function that "cultivates and implements a culture of risk management" in organizations. It helps the other functions, "incorporates processes to assess potential impacts", and connects AI system work to "organizational values and principles". GOVERN has six categories and 19 subcategories. Its outcomes are voluntary guidance. See How Govern fits into NIST AI RMF 1.0.

What are the six categories of the NIST AI RMF Govern function?

NIST AI 100-1, Table 1, gives these six GOVERN categories:

• GOVERN 1: policies, processes, procedures, and practices for AI risk management.
• GOVERN 2: accountability structures.
• GOVERN 3: workforce diversity, equity, inclusion, and accessibility.
• GOVERN 4: "a culture that considers and communicates AI risk".
• GOVERN 5: "robust engagement with relevant AI actors".
• GOVERN 6: "AI risks and benefits arising from third-party software and data and other supply chain issues".

As of October 2026, NIST reports that AI RMF 1.0 "is being revised as part of the White House AI Action Plan", with no new version available. GOVERN 3 continues to use the January 2023 text. See The six GOVERN categories.

How many subcategories does the NIST AI RMF Govern function have?

GOVERN has 19 subcategories in NIST AI 100-1, Table 1. The counts are GOVERN 1: seven, GOVERN 2: three, GOVERN 3: two, GOVERN 4: three, GOVERN 5: two, and GOVERN 6: two. Each subcategory gives an outcome, and organizations can select categories and subcategories or apply all of them. See The six GOVERN categories.

How does Govern relate to Map, Measure, and Manage in NIST AI RMF?

GOVERN helps all three other functions during NIST AI RMF risk management. NIST recommends that "Aspects of GOVERN, especially those related to compliance or evaluation, should be integrated into each of the other functions". GOVERN applies to "all stages". MAP, MEASURE, and MANAGE "can be applied in AI system-specific contexts and at specific stages of the AI lifecycle". After GOVERN, most users start with MAP.

If there is a governance structure, the functions can occur in a different sequence. NIST recommends that "the process should be iterative". See How Govern fits into NIST AI RMF 1.0 and the sequence conditions.

What is the NIST AI RMF Map function?

MAP gives context for risks related to an AI system in NIST AI RMF 1.0. Its five categories include context, categorization, capabilities and benefits and costs, component risks and benefits, and impacts. Its outcomes "are the basis for the MEASURE and MANAGE functions". The framework gives voluntary guidance and no obligations.

As of October 2026, NIST reports that AI RMF 1.0 "is being revised as part of the White House AI Action Plan", but January 2023 is the last released version. See How Map fits into NIST AI RMF 1.0.

What are the five categories of the NIST AI RMF Map function?

NIST AI 100-1, Table 2, gives five MAP categories:

• MAP 1: context.
• MAP 2: categorization of the AI system.
• MAP 3: "AI capabilities, targeted usage, goals, and expected benefits and costs compared with appropriate benchmarks".
• MAP 4: risks and benefits of all AI system components, with third-party software and data.
• MAP 5: impacts to individuals, groups, communities, organizations, and society.

Each category has subcategories that give outcomes. See The five MAP categories.

How many subcategories does the NIST AI RMF Map function have?

MAP has 18 subcategories in NIST AI 100-1, Table 2. The counts are MAP 1: six, MAP 2: three, MAP 3: five, MAP 4: two, and MAP 5: two. Each gives an outcome. See The five MAP categories.

When should the NIST AI RMF Map function be applied in the AI lifecycle?

NIST AI 100-1 sets no one lifecycle point for MAP. After GOVERN, most users start with MAP, but if there is a governance structure, a different sequence is possible. NIST recommends that "the process should be iterative". Section 5.2 recommends sufficient context for an initial decision "whether to design, develop, or deploy an AI system".

Framework users continue MAP as "context, capabilities, risks, benefits, and potential impacts" change. This is voluntary guidance. See How Map fits into NIST AI RMF 1.0.

How is Map different from Govern in NIST AI RMF?

MAP gives context for AI system risks, while GOVERN helps all NIST AI RMF functions during the risk management process. Governance is continual across "an AI system's lifespan and the organization's hierarchy". MAP, MEASURE, and MANAGE "can be applied in AI system-specific contexts and at specific stages of the AI lifecycle". MAP 3.5 connects human oversight to the policies from GOVERN. See How Map fits into NIST AI RMF 1.0, the GOVERN function, and MAP 3.5.

What is the NIST AI RMF Measure function?

MEASURE uses quantitative, qualitative, or mixed methods "to analyze, assess, benchmark, and monitor AI risk and related impacts". In NIST AI 100-1, Section 5.3, it uses information from MAP and "informs the MANAGE function". NIST recommends that "AI systems should be tested before their deployment and regularly while in operation". The metrics include trustworthy characteristics, "social impact, and human-AI configurations". Its four categories and 22 subcategories give voluntary outcomes. See How Measure fits into NIST AI RMF 1.0.

What are the four categories of the NIST AI RMF Measure function?

NIST AI 100-1, Table 3, gives four MEASURE categories:

• MEASURE 1: "Appropriate methods and metrics are identified and applied."
• MEASURE 2: "AI systems are evaluated for trustworthy characteristics."
• MEASURE 3: "Mechanisms for tracking identified AI risks over time are in place."
• MEASURE 4: "Feedback about efficacy of measurement is gathered and assessed."

These categories hold 22 subcategories. See The four MEASURE categories.

How many subcategories does the NIST AI RMF Measure function have?

MEASURE has 22 subcategories in NIST AI 100-1, Table 3. MEASURE 1, 3, and 4 each have three, while MEASURE 2 has 13. MEASURE 2 has more subcategories than each other Core category. See The four MEASURE categories.

Why does MEASURE 2 have so many subcategories?

MEASURE 2 has 13 subcategories for the evaluation of AI systems for trustworthy characteristics (NIST AI 100-1, Table 3). On the Trustworthy AI page, we read MEASURE 2.5 to 2.11 as the subcategories most directly related to the seven characteristics. NIST gives no one-to-one mapping: MEASURE 2.6 safety metrics also "reflect system reliability and robustness".

The other subcategories include:

• TEVV documentation (2.1), "Evaluations involving human subjects" (2.2), "performance or assurance criteria" (2.3), and monitoring "when in production" (2.4).
• "Environmental impact and sustainability" (2.12), which is not one of the seven characteristics.
• "Effectiveness of the employed TEVV metrics and processes" (2.13).

See MEASURE 2 and the characteristic mapping.

How does Measure relate to Map and Manage in NIST AI RMF?

MEASURE uses information from MAP and gives MANAGE measurement outcomes for risk monitoring and response (NIST AI 100-1, Section 5.3). After GOVERN, most users start with MAP.

If there is a governance structure, the functions can occur in a different sequence. NIST recommends that "the process should be iterative". GOVERN applies to "all stages", while the other functions "can be applied in AI system-specific contexts and at specific stages of the AI lifecycle".

Framework users continue MEASURE as "knowledge, methodologies, risks, and impacts" change. See How Measure fits into NIST AI RMF 1.0 and the sequence conditions.

What is the NIST AI RMF Manage function?

MANAGE "entails allocating risk resources to mapped and measured risks on a regular basis and as defined by the GOVERN function" (NIST AI 100-1, Section 5.4). Risk treatment includes "plans to respond to, recover from, and communicate about incidents or events". Its four categories and 13 subcategories give voluntary outcomes.

More conditions are:

• The responses in MANAGE 1.3 are examples, not the only responses.
• MANAGE 1.4 has this outcome: "Negative residual risks (defined as the sum of all unmitigated risks) to both downstream acquirers of AI systems and end users are documented." It has no outcome to give the documentation to other persons or to accept a residual risk.
• Section 1.2.3 gives the effect of residual risk documentation: information for end users about "potential negative impacts".

See How Manage fits into NIST AI RMF 1.0.

What are the four categories of the NIST AI RMF Manage function?

NIST AI 100-1, Table 4, gives four MANAGE categories:

• MANAGE 1: AI risks from MAP and MEASURE are "prioritized, responded to, and managed".
• MANAGE 2: "Strategies to maximize AI benefits and minimize negative impacts", with input from "relevant AI actors".
• MANAGE 3: AI risks and benefits from third-party entities.
• MANAGE 4: documented "Risk treatments, including response and recovery, and communication plans", with regular monitoring.

These categories hold 13 subcategories. See The four MANAGE categories.

How many subcategories does the NIST AI RMF Manage function have?

MANAGE has 13 subcategories in NIST AI 100-1, Table 4. MANAGE 1 and 2 each have four, MANAGE 3 has two, and MANAGE 4 has three. Each gives a Core outcome. The full Core has 72 subcategories. See The four MANAGE categories.

What risk response options does NIST AI RMF identify?

NIST AI 100-1, MANAGE 1.3, gives examples: "Risk response options can include mitigating, transferring, avoiding, or accepting". These are examples, not the only responses or obligations. MANAGE 1.3 is about documented responses to AI risks "deemed high priority", as MAP identifies them.

More conditions are:

• "impact, likelihood, and available resources or methods" govern prioritization in MANAGE 1.2.
• "Negative residual risks (defined as the sum of all unmitigated risks)" to "downstream acquirers" and end users have documentation as the outcome in MANAGE 1.4.
• Modulos records response and residual risk documentation with the Manage 1.3 and Manage 1.4 Requirements. The risk register does not record the selected response or if the organization accepts a residual risk.

See MANAGE 1.3.

How does Manage relate to Map and Measure in NIST AI RMF?

MANAGE uses MAP and MEASURE output for risk management. This is "on a regular basis and as defined by the GOVERN function" (NIST AI 100-1, Section 5.4). MAP supplies the context, and measurement outcomes help with risk monitoring and response. After GOVERN, most users start with MAP.

If there is a governance structure, the functions can occur in a different sequence. NIST recommends that "the process should be iterative". Framework users continue MANAGE for deployed AI systems as "methods, contexts, risks, and needs or expectations from relevant AI actors" change. See How Manage fits into NIST AI RMF 1.0.

What are the 12 generative AI risk categories defined in NIST AI 600-1?

NIST AI 600-1 names them risks, and Section 2 gives these 12:

1. "CBRN Information or Capabilities"
2. "Confabulation"
3. "Dangerous, Violent, or Hateful Content"
4. "Data Privacy"
5. "Environmental Impacts"
6. "Harmful Bias or Homogenization"
7. "Human-AI Configuration"
8. "Information Integrity"
9. "Information Security"
10. "Intellectual Property"
11. "Obscene, Degrading, and/or Abusive Content"
12. "Value Chain and Component Integration"

Section 2.6 uses "Harmful Bias and Homogenization" for risk 6. The risks "can be further categorized by organizations" (footnote 5). The Profile "focuses on risks for which there is an existing empirical evidence base at the time this profile was written". See The 12 risks, with the related trustworthy AI characteristics in Sections 2.1 to 2.12.

When was the NIST Generative AI Profile published?

NIST released AI 600-1 on July 26, 2024, as a companion resource to AI RMF 1.0 of January 2023. It "was developed per Section 4.1(a)(i)(A) of EO 14110", which Executive Order 14148 revoked on January 20, 2025 (Section 2(ggg)). The work used input from the Generative AI Public Working Group. Section 1 gives this information: "Future revisions of this profile will include additional AI RMF subcategories, risks, and suggested actions". See the Profile's date and source and its structure.

How does NIST AI 600-1 relate to NIST AI RMF 1.0 (AI 100-1)?

NIST AI 600-1 applies AI RMF 1.0 to generative AI as a cross-sectoral profile and companion resource, not a replacement. Profiles are "based on the requirements, risk tolerance, and resources of the Framework user". Suggested actions apply to only some Core subcategories, with Action IDs and risk tags, for example GV-1.1-001 for GOVERN 1.1.

AI RMF 1.0 and Playbook actions also apply to GAI risks. These documents are voluntary guidance. See How the Generative AI Profile fits into NIST AI RMF 1.0.

What is the relationship between NIST AI 600-1 and the EU AI Act GPAI obligations?

NIST AI 600-1 is voluntary guidance, while the EU AI Act is law. The two documents do not refer to each other. A general-purpose AI model in Article 3(63) is different from GAI in AI 600-1. Chapter V gives obligations to providers of general-purpose AI models. Article 53 sets obligations such as a copyright policy, with no risk assessment obligation. Article 55(1) adds obligations for providers of models with systemic risk, for example to "assess and mitigate possible systemic risks at Union level".

Article 50(2) and 50(4) give transparency obligations to providers and deployers of some AI systems that make synthetic content, with exceptions. For these provisions and conditions, see Generative AI in other laws and standards.

What are the seven characteristics of trustworthy AI in NIST AI RMF 1.0?

NIST AI 100-1, Section 3, gives seven characteristics:

• valid and reliable
• safe
• secure and resilient
• accountable and transparent
• explainable and interpretable
• privacy-enhanced
• fair – with harmful bias managed.

NIST recommends a balance that uses the context of use. The framework is voluntary guidance.

As of October 2026, NIST reports that AI RMF 1.0 "is being revised as part of the White House AI Action Plan", but no changed version is available. The characteristics are those of January 2023. See The seven characteristics and revision status.

Why is 'Valid and Reliable' the base characteristic in NIST AI RMF?

NIST AI 100-1, Figure 4, gives valid and reliable as "a necessary condition of trustworthiness" and shows it "as the base for other trustworthiness characteristics". Figure 4 shows accountable and transparent "as a vertical box because it relates to all other characteristics". Figure 4 sets no sequence for the other five characteristics. Section 3.1 tells organizations that systems "which are inaccurate, unreliable, or poorly generalized" increase risks and decrease trustworthiness. See Figure 4 and the tradeoffs between the characteristics.

How do trustworthy AI characteristics relate to the four NIST AI RMF functions (Govern, Map, Measure, Manage)?

Some NIST AI RMF Core outcomes use the characteristics: GOVERN 1.2 makes them part of "organizational policies, processes, procedures, and practices", and MEASURE 2 is their evaluation in AI systems. On the Trustworthy AI page, we read each of MEASURE 2.5 to 2.11 as most directly related to one characteristic. NIST gives no one-to-one mapping: MEASURE 2.6 safety metrics also show reliability and robustness.

That table gives no characteristic for MEASURE 2.1 to 2.4, 2.12, or 2.13. MEASURE also includes "social impact, and human-AI configurations". See How the trustworthy AI characteristics fit into NIST AI RMF 1.0.

Can trustworthy AI characteristics conflict with each other?

Yes. NIST AI 100-1, Section 3, tells organizations that "Addressing AI trustworthiness characteristics individually will not ensure AI system trustworthiness". Tradeoffs are usual, and the context changes which characteristics apply and which are more important. Examples include interpretability and privacy, accuracy and interpretability, and decreased accuracy with privacy methods where there is "data sparsity".

The decision uses the context and its values. NIST recommends that tradeoffs "should be resolved in a manner that is both transparent and appropriately justifiable". The evaluation of a tradeoff does not give the decision. This is voluntary guidance. See Figure 4 and the tradeoffs between the characteristics.

How do the NIST trustworthy AI characteristics map to the EU AI Act and ISO/IEC 42001?

NIST AI 100-1 does not map its characteristics to Regulation (EU) 2024/1689 or ISO/IEC 42001. The crosswalks page of the AI Resource Center of NIST shows a document of January 26, 2023: "An illustration of how NIST AI RMF trustworthiness characteristics relate to the OECD Recommendation on AI, Proposed EU AI Act, Executive Order 13960, and Blueprint for an AI Bill of Rights". This text names the Proposed EU AI Act, not Regulation (EU) 2024/1689 (sources of the characteristics page). The AI RMF is voluntary and "law- and regulation-agnostic". The EU AI Act sets requirements for high-risk AI systems in Articles 8 to 15, while ISO/IEC 42001 sets AIMS requirements.

The ISO/IEC crosswalk that NIST shows maps Core subcategories, not characteristics, to clauses and Annex B. See How NIST AI RMF compares to other frameworks.