NIST AI RMF: what is it and who needs it?

The NIST AI Risk Management Framework is voluntary guidance for managing risks that AI systems pose to people, organizations and society. No one is certified against it. Teams that build, buy or deploy AI, and customers who ask for evidence of responsible practice, use it as a shared structure.

This explainer covers the framework and how to start. It does not claim any Safeguard certification, and Safeguard does not certify your organization.

What is it, exactly?

NIST describes the AI RMF as a framework developed to better manage risks to individuals, organizations and society associated with AI. It is intended for voluntary use and aims to help organizations build trustworthiness into the design, development, use and evaluation of AI products, services and systems.

NIST released version 1.0 on 26 January 2023, after a consensus-driven process with several public drafts and workshops. A companion Playbook, a Roadmap and a Crosswalk were published alongside it. On 26 July 2024 NIST released NIST AI 600-1, the Generative Artificial Intelligence Profile, which helps organizations identify risks unique to generative AI and proposes actions for managing them.

Two current notes from NIST's page, as read on 18 October 2026: the AI RMF 1.0 is being revised as part of the White House AI Action Plan, and on 7 April 2026 NIST released a concept note for a profile on trustworthy AI in critical infrastructure. Check the page for the latest status before citing a version.

Who needs it?

Nobody is required to follow it by the framework itself, since it is voluntary. In practice three groups use it.

Builders of AI features want a way to show customers they thought about risk. Buyers use it as a questionnaire backbone for vendors. Regulated and public-sector organizations may find it referenced in their own policies or contracts, so read your contracts rather than assuming.

If you only use a vendor's AI tool, the framework still helps you ask good questions: what data goes in, who can see outputs and what happens when the model is wrong.

How is it organized?

The framework is built around outcomes rather than a checklist. NIST's AI RMF Playbook and Resource Center offer suggested actions for each. We did not fetch the framework text itself for this post, so read the core document for the exact function names and categories before you map controls. What the NIST overview page does tell us is that it is meant to be adapted to your goals and priorities, and the generative AI profile does the same for its own risks.

How does it relate to other work you already do?

Security teams often have SSDF or ISO 27001 programs. NIST also publishes SP 800-218A, an SSDF community profile for secure development of generative AI and dual-use foundation models, which adds practices specific to model development. The AI RMF is broader than security, covering fairness, explainability and misuse.

Treat the AI RMF as the umbrella for governance and your security frameworks as the technical controls beneath it.

What to do

1. List every AI system you build or use, including vendor features and shadow tools.
2. Name an owner for each system and record its purpose and data.
3. Read the AI RMF 1.0 and the generative AI profile, AI 600-1, if you use generative models.
4. Use the Playbook to pick a handful of actions per system, rather than trying all at once.
5. Record decisions and evidence as you go, so a customer questionnaire becomes a copy exercise.
6. Add model and dataset provenance to your supply chain inventory.
7. Revisit when NIST publishes the revision or new profiles.

Quick check

Sources

• NIST AI Risk Management Framework overview, fetched 18 October 2026
• NIST Secure Software Development Framework project page, fetched 18 October 2026

Last reviewed: 18 October 2026.