NIST AI Risk Management Framework Guide (2026): Govern, Map, Measure, Manage

Who it applies to

The AI RMF is "voluntary, rights-preserving, non-sector-specific, and use-case agnostic" (NIST AI 100-1). It is written for AI actors across the AI lifecycle, as defined by the OECD (NIST AI 100-1). They include:

• people who design, develop and deploy AI
• people who operate, test, evaluate, verify and validate AI
• affected communities and other interested parties, who inform the primary audience

Typical users are:

• AI product companies that want to show trustworthy-AI practice to enterprise buyers
• enterprises deploying third-party AI
• organisations that align their AI policies to US federal expectations

Key requirements

Trustworthy AI characteristics

NIST AI 100-1 lists seven characteristics (NIST AI 100-1):

1. valid and reliable
2. safe
3. secure and resilient
4. accountable and transparent
5. explainable and interpretable
6. privacy-enhanced
7. fair, with harmful bias managed

Valid and reliable is described as "a necessary condition of trustworthiness." Accountable and transparent relates to all the other characteristics (NIST AI 100-1).

The Core: four functions and 19 categories

From the Core tables in AI 100-1 (NIST AI 100-1):

• GOVERN (6 categories): policies, processes and practices, starting with legal and regulatory requirements (GOVERN 1.1). Also covers roles and responsibilities, workforce diversity in decision-making, risk culture, engagement with AI actors, and third-party software, data and supply chain risk (GOVERN 6).
• MAP (5 categories): establishes context, including intended purposes and uses (MAP 1.1). Also covers categorising the AI system, benefits and costs, risks from third-party components, and the likelihood and magnitude of impacts (MAP 5).
• MEASURE (4 categories): chooses metrics and methods (MEASURE 1). Evaluates the system for trustworthy characteristics through test sets and tools (MEASURE 2), tracks risks over time (MEASURE 3), and gathers feedback on how well measurement is working (MEASURE 4).
• MANAGE (4 categories): decides whether the AI system should proceed (MANAGE 1.1). Also covers allocating resources to manage risk, managing third-party risks and benefits (MANAGE 3.1), and post-deployment monitoring and response (MANAGE 4.1).

Profiles

AI RMF profiles come in three types (NIST AI 100-1):

• use-case profiles, which implement the functions for a specific setting
• temporal profiles, a Current Profile and a Target Profile used for gap analysis
• cross-sectoral profiles, which cover risks that apply across sectors

Generative AI Profile (NIST AI 600-1)

The profile defines risks that are "novel to or exacerbated by" generative AI (NIST AI 600-1):

1. CBRN information or capabilities
2. confabulation
3. dangerous, violent or hateful content
4. data privacy
5. environmental impacts
6. harmful bias or homogenisation
7. human-AI configuration
8. information integrity
9. information security
10. intellectual property
11. obscene, degrading or abusive content
12. value chain and component integration

It then gives suggested actions for managing these risks (NIST AI 600-1). NIST's publication page says the profile was prepared pursuant to EO 14110 (NIST).

Certification process and timeline

There is no official certification or attestation against the AI RMF. A typical adoption path:

1. Build a Current Profile of how AI is governed and used today (NIST AI 100-1).
2. Define a Target Profile based on your risk tolerance and legal requirements.
3. Prioritise the gaps. Start with GOVERN, because it is cross-cutting.
4. Use the Playbook and crosswalk documents published through the Trustworthy and Responsible AI Resource Center to choose actions (NIST; NIST AIRC).
5. Review the framework periodically. NIST encourages organisations to evaluate it regularly (NIST AI 100-1).

Version timeline:

• AI RMF 1.0: January 2023.
• GenAI Profile: 26 July 2024.
• Critical Infrastructure Profile concept note: 7 April 2026.
• Revision: in progress (NIST). AI 100-1 also commits to a review with formal community input "no later than 2028" (NIST AI 100-1).

Costs

The framework, Playbook and profiles are free to download from NIST (NIST). No implementation cost figures are cited.

Penalties

The AI RMF is voluntary and carries no penalties (NIST). Obligations come from the laws and contracts you map to it, which GOVERN 1.1 explicitly asks you to identify (NIST AI 100-1).

Mapping to related frameworks

• ISO/IEC 42001: a certifiable AI management system. The AI RMF's GOVERN, MAP, MEASURE and MANAGE activities can provide the risk-process content.
• EU AI Act: use MAP to inventory and categorise systems, and record legal obligations under GOVERN 1.1.
• NIST CSF 2.0 and NIST Privacy Framework: these cover the "secure and resilient" and "privacy-enhanced" characteristics.
• NIST crosswalks: official crosswalk documents are published through the AIRC (NIST AIRC).

Readiness checklist

1. Make an inventory of the AI systems you develop, buy or embed, with their intended purposes (MAP 1.1).
2. Record the applicable laws and regulations (GOVERN 1.1).
3. Assign roles, responsibilities and lines of communication (GOVERN 2).
4. Assess third-party AI, data and supply-chain risk (GOVERN 6, MANAGE 3).
5. Define metrics and test sets for each trustworthy characteristic (MEASURE 1–2).
6. For generative AI, assess the 12 risks in NIST AI 600-1.
7. Plan post-deployment monitoring and incident response (MANAGE 4).
8. Document Current and Target Profiles and review them on a regular schedule.

Common pitfalls

• Treating GOVERN as a one-time policy instead of a cross-cutting function.
• Measuring only accuracy. NIST lists seven characteristics.
• Ignoring third-party models, which are covered explicitly in GOVERN 6 and MANAGE 3.
• Claiming "AI RMF certified." No such certification exists.
• Not tracking the 2026 revision of AI RMF 1.0 (NIST).

How AssureCore helps

• The Risk Register records AI risks with likelihood, impact and treatment.
• Cross-framework Mapping links AI RMF activities to ISO/IEC 42001 and the EU AI Act.
• Policy Management drafts AI governance policies with AI-assisted writing and gap analysis.
• Evidence Management keeps evidence for tests and monitoring linked to controls, with expiry alerts.

Frequently asked questions

Official references

• NIST – AI Risk Management Framework
• NIST AI 100-1 – AI RMF 1.0 (PDF)
• NIST AIRC – AI RMF resources
• NIST – AI 600-1 Generative AI Profile publication page
• NIST AI 600-1 – Generative AI Profile (PDF)

Related frameworks

• EU AI Act guide
• ISO 42001 guide
• UK AI Reg guide
• NIST CSF guide