What is the NIST AI Risk Management Framework, and is it mandatory?
By Nick Jang, founder of MINERVA Studio ·
The short answer
The NIST AI RMF is the Artificial Intelligence Risk Management Framework published by the US National Institute of Standards and Technology. Version 1.0 was released on 26 January 2023. It is voluntary: a structure, not a rule, organized around four functions, Govern, Map, Measure and Manage, that any organization can use to manage the risks of the AI it builds or uses.
What the letters stand for
NIST is the National Institute of Standards and Technology, an agency of the US Department of Commerce. AI RMF stands for Artificial Intelligence Risk Management Framework. Congress asked NIST to write it in the National AI Initiative Act of 2020, and NIST developed it in the open, with drafts and workshops, before publishing version 1.0 (NIST AI 100-1).
The four functions
| Function | What it means | Examples of what to do |
|---|---|---|
| Govern | A culture and structure for AI risk, across the organization. It applies to the other three. | Policies, roles and accountability, risk tolerance, training, third-party AI policies. |
| Map | Understand the context of each AI system and the risks it creates. | Intended purpose, users and affected people, limits of the method, impacts. |
| Measure | Analyse, assess and track the risks that were mapped. | Testing for accuracy, robustness, bias, security and privacy; metrics over time. |
| Manage | Act on the risks: prioritize, treat, monitor and respond. | Decide to deploy or not, mitigations, incident response, decommissioning. |
Each function breaks into categories and subcategories. They are outcomes to aim for, not a checklist; the framework expects organizations to choose what fits.
The seven characteristics of trustworthy AI
The framework describes trustworthy AI as having seven characteristics, and treats managing AI risk as balancing them:
- Valid and reliable
- Safe
- Secure and resilient
- Accountable and transparent
- Explainable and interpretable
- Privacy-enhanced
- Fair, with harmful bias managed
The Playbook and the Generative AI Profile
Two companions make the framework usable. The AI RMF Playbook suggests actions, references and documentation for each subcategory. The Generative AI Profile (NIST AI 600-1), published in July 2024, applies the framework to generative AI and lists twelve risks that generative models create or make worse, among them confabulation (confident wrong answers), information security, harmful bias, data privacy and intellectual property, with suggested actions for each.
Is the NIST AI RMF mandatory?
No, and NIST says so in the framework itself. Nothing in it is enforced against private companies. It only becomes required when someone makes it so: a customer contract, a sector regulator that points at it, a government policy for its own agencies. OSFI doesn’t require it of Canadian banks; E-23 and the AI bulletin set OSFI’s expectations. Plenty of institutions use the NIST structure anyway, because it’s clear, free, and auditors and vendors already know their way around it.
Using the NIST AI RMF beside OSFI E-23
| NIST AI RMF | Where E-23 asks for the same |
|---|---|
| Govern | Governance outcome: accountability, the model risk framework tied to the risk appetite, enough of the right people. |
| Map | Model purpose, scope and use stated by the owner; risk rating; the inventory. |
| Measure | Development testing, independent review, monitoring against thresholds. |
| Manage | Approval with mitigants, escalation of breaches, decommissioning. |
MINERVA ships an AI risk rule pack built on the NIST AI RMF beside the E-23 pack, so one inventory can be read against both. It doesn’t certify conformity to either.
Related questions
When was the NIST AI RMF released?
Version 1.0 was released on 26 January 2023. The Generative AI Profile (NIST AI 600-1) followed in July 2024.
Can you be certified against the NIST AI RMF?
No. There is no NIST certification. Organizations that want a certifiable standard look to ISO/IEC 42001.
What is the difference between the NIST AI RMF and ISO 42001?
The NIST AI RMF is a voluntary risk framework. ISO/IEC 42001 is a management system standard with requirements an accredited body can certify. Many organizations use NIST for risk work and ISO for the management system around it.
Does the NIST AI RMF apply in Canada?
It has no legal force in Canada, but Canadian organizations can and do use it as a voluntary framework.
Sources
- NIST — AI Risk Management Framework
- NIST — AI RMF 1.0 (NIST AI 100-1)
- NIST — Generative AI Profile (NIST AI 600-1)
New answers are posted on MINERVA on LinkedIn. Follow along, or send the question you’d like answered next to minerva@minerva.studio.
A plain-language explanation, not legal advice. The regulator’s own text is the source; read it, and take advice on how it applies to your institution.