NIST AI Risk Management Framework (AI RMF)
A voluntary framework from the US National Institute of Standards and Technology for managing the
risks of AI systems. AI RMF 1.0 (NIST AI 100-1) was released on 2023-01-26. It is not law and not
certifiable, but it is widely used as a shared risk vocabulary, also outside the US. NIST states
that AI RMF 1.0 “is being revised as part of the White House AI Action Plan”; no revised text was
published as of 2026-11-10. Overview of governance frameworks: ai-safety-and-governance.
Structure
The framework has two parts.
- Foundational information: how to frame AI risk, who the AI actors are, and what makes AI
trustworthy. - The Core: four functions, each split into categories and subcategories (outcomes to reach).
| Function | Purpose |
|---|---|
| Govern | Culture, policies, roles and accountability for AI risk across the organisation. Applies to the other three |
| Map | Establish context: purpose, users, impacts, and the risks of a specific AI system |
| Measure | Analyse, assess and track the mapped risks with quantitative and qualitative methods |
| Manage | Prioritise risks, act on them, and monitor and respond over time |
Trustworthiness characteristics
- Valid and reliable
- Safe
- Secure and resilient
- Accountable and transparent
- Explainable and interpretable
- Privacy-enhanced
- Fair, with harmful bias managed
Valid and reliable is the base for the others; accountable and transparent cuts across all of them.
Companion resources
- AI RMF Playbook: suggested actions,
documentation and references for every subcategory. NIST says it is “neither a checklist nor set
of steps to be followed in its entirety”. It can be downloaded as PDF, CSV, XLSX or JSON, and is
to be updated after the framework revision. - Generative AI Profile (NIST AI 600-1), released 2024-07-26: the risks specific to or made
worse by generative AI (such as confabulation, information integrity, information security,
harmful bias, intellectual property, data privacy) with suggested actions per function. - Profiles: a profile applies the Core to one sector or use case. A concept note for a critical
infrastructure profile was published on 2026-04-07. - Crosswalks: mappings to other frameworks,
including ISO/IEC 42001 (by Microsoft), ISO/IEC 23894 (revised, 2025-08-14), ISO/IEC 42005
(2025-08-14), the OECD recommendation and the proposed EU AI Act (2023-01-26), Singapore’s AI
Verify, Korea’s TTA guidebook and Japan’s AI Guidelines for Business. - Trustworthy and Responsible AI Resource Center (AIRC), launched
2023-03-30, hosts the framework, the Playbook and the crosswalks.
How it relates to the others
- IEC 42001 is a certifiable management system: it says what an
organisation must have in place. The AI RMF says which risk outcomes to aim for and suggests
actions. Teams often use the RMF Playbook to fill in ISO 42001’s risk and impact assessment
steps. - EU AI Act: Map and Measure line up with the Act’s risk management, data
governance and accuracy requirements (Arts. 9, 10 and 15); Govern lines up with the quality
management system and human oversight duties. The NIST crosswalk to the Act dates from the 2023
proposal, not the final text. - In practice: run evaluations and red teaming (llm-as-judge-and-evals,
agent-evals-and-observability) as Measure activities, and defend against prompt injection
(prompt-injection-and-agent-security) under Manage.
Sources (accessed 2026-11-10)
- NIST AI RMF page: https://www.nist.gov/itl/ai-risk-management-framework
- AIRC, AI RMF overview: https://airc.nist.gov/airmf-resources/airmf/
- AIRC, trustworthiness characteristics: https://airc.nist.gov/AI_RMF_Knowledge_Base/AI_RMF/Foundational_Information/3-sec-characteristics
- AIRC, Playbook: https://airc.nist.gov/airmf-resources/playbook/
- AIRC, crosswalks: https://airc.nist.gov/airmf-resources/crosswalks/