The NIST AI Risk Management Framework: Govern, Map, Measure, Manage

The AI Risk Management Framework (AI RMF) is published by the US National Institute of Standards and Technology. Version 1.0, document NIST AI 100-1, was released on 26 January 2023. Unlike the EU AI Act, it is not a law: it is a voluntary framework that organisations can adopt to manage the risks of AI systems they build or use.

Seven characteristics of trustworthy AI

The framework describes what it means for an AI system to be trustworthy through seven characteristics:

1. Valid and reliable - it does what it is meant to do, consistently.
2. Safe - it does not endanger human life, health, property or the environment.
3. Secure and resilient - it withstands attacks and recovers from unexpected events.
4. Accountable and transparent - information about the system is available to the people who interact with it.
5. Explainable and interpretable - how it works, and what an output means, can be understood.
6. Privacy-enhanced - it protects people's autonomy and personal data.
7. Fair, with harmful bias managed - it addresses equality and equity concerns.

The framework treats these as trade-offs to be balanced rather than boxes to tick: improving one, such as interpretability, can come at the cost of another, such as accuracy.

The four core functions

The practical part of the framework is organised as four functions:

• Govern - build a culture and structure for risk management: policies, accountability, roles and processes. NIST describes it as cross-cutting, applying throughout the other three.
• Map - establish context: what the system is for, who it affects, and what could go wrong.
• Measure - analyse, assess and track the risks identified, using quantitative and qualitative methods.
• Manage - prioritise the risks and act on them, including deciding whether a system should be deployed at all.

Each function is broken into categories and subcategories of outcomes. NIST publishes a companion Playbook with suggested actions for each.

The Generative AI Profile

On 26 July 2024 NIST released NIST AI 600-1, the Generative AI Profile. It applies the framework to systems such as the language models covered on this site, and lists 12 risks that are unique to generative AI or made worse by it:

• CBRN information or capabilities
• Confabulation (confidently stated false output)
• Dangerous, violent or hateful content
• Data privacy
• Environmental impacts
• Harmful bias and homogenisation
• Human-AI configuration
• Information integrity
• Information security
• Intellectual property
• Obscene, degrading or abusive content
• Value chain and component integration

For each risk, the profile maps suggested actions to the four functions above.

Recent updates

The NIST page lists further work since then. On 7 April 2026 NIST released a concept note for an AI RMF Profile on Trustworthy AI in Critical Infrastructure, aimed at operators of critical infrastructure using AI. NIST also states that AI RMF 1.0 is being revised as part of the White House AI Action Plan; the page gives no date for the revised version.

Related

• AI Ethics: Understanding Bias, Safety and Responsible AI
• How to read an AI model card - where vendors publish the evaluation results this kind of framework asks for.