Lesson 1

What the AI RMF is, and what it is not

Video
2 min
Reading
10 min
Exercise
7 min
Total
18 min

Educational content, not legal advice.

In this lesson you will

  1. Explain what the NIST AI Risk Management Framework is for, and that it is voluntary.
  2. Describe its two parts: the trustworthiness characteristics and the four functions of the Core.
  3. Say what "aligned with NIST" can honestly mean, and write it down before using it.

Video lesson

Length: 1:07
Read the transcript

This is lesson one of NIST AI RMF in practice. We follow Tessmoor Credit Union, our fictional credit union, whose board has asked its risk team to adopt the NIST framework.

NIST released the AI Risk Management Framework in January 2023, and states that it is intended for voluntary use. NIST AI RMF 1.0 (NIST AI 100-1) It is a framework, not law, and NIST says it is being revised, so name the version you use. NIST AI RMF 1.0 (NIST AI 100-1)

The framework has two parts. First, seven characteristics of trustworthy AI, from valid and reliable to fair with harmful bias managed. NIST AI 100-1 (AI RMF 1.0), Part 1, section 3 and Figure 4 Second, the Core: four functions, GOVERN, MAP, MEASURE and MANAGE. NIST AI 100-1 (AI RMF 1.0), Executive Summary; Part 2, section 5 (AI RMF Core) They are not a checklist, and GOVERN runs through the other three. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5

The rules that bind come first. Generative AI does not sit outside Canadian privacy law. OPC and FPT privacy authorities, Principles for responsible, trustworthy and privacy-protective generative AI technologies (December 7, 2023)

Recap: voluntary, two parts, four functions, and an honest claim names its scope and its evidence. Next: GOVERN.

Notes

Why people reach for NIST

Sooner or later someone asks an organization to "use NIST" for its AI. A customer puts it in a questionnaire, a board member reads about it, a vendor says its product is aligned with it. Before anyone answers yes, it helps to know what the framework is, what it asks for, and what it leaves to you.

Meet Tessmoor Credit Union

Tessmoor Credit Union is a fictional Canadian credit union. It uses two AI systems:

  • a model that scores small business loan applications, which an underwriter reads before deciding;
  • a generative AI assistant in its mobile app that answers members' questions about accounts and products.

Its board has asked the risk team to "adopt the NIST framework" for both. This course follows Tessmoor's AI risk lead as they work out what that means, one function at a time.

What NIST released, and its status

NIST released the AI Risk Management Framework, AI RMF 1.0 (NIST AI 100-1), on January 26, 2023, and states that it is intended for voluntary use. NIST AI RMF 1.0 (NIST AI 100-1) It is described as voluntary, rights preserving, not specific to any sector and agnostic about use cases; it is a framework, not law. NIST AI 100-1 (AI RMF 1.0), Executive Summary; Part 2, section 5 (AI RMF Core) NIST also states that AI RMF 1.0 is being revised as part of the White House AI Action Plan. NIST AI RMF 1.0 (NIST AI 100-1) So every time this course names a function or category, it means AI RMF 1.0, and anyone using the framework checks the NIST page for a new version before each review.

What the framework means by risk

In the AI RMF, risk is the composite measure of an event's probability of occurring and the magnitude of its consequences. NIST AI 100-1 (AI RMF 1.0), Part 1, section 1.1 The impacts of AI systems can be positive, negative or both, and harm can be felt by individuals, groups, communities, organizations, society, the environment and the planet. NIST AI 100-1 (AI RMF 1.0), Part 1, section 1.1 That wide circle matters for Tessmoor: a loan model that is accurate on average can still harm one group of applicants.

Trustworthy AI: seven characteristics

The framework lists the characteristics of trustworthy AI systems: valid and reliable; safe; secure and resilient; accountable and transparent; explainable and interpretable; privacy enhanced; and fair with harmful bias managed. NIST AI 100-1 (AI RMF 1.0), Part 1, section 3 and Figure 4 Valid and reliable is the base the others rest on, and accountable and transparent relates to all the others. NIST AI 100-1 (AI RMF 1.0), Part 1, section 3 and Figure 4 The characteristics are balanced against each other in the context of use, and tradeoffs are usually involved. NIST AI 100-1 (AI RMF 1.0), Part 1, section 3 and Figure 4 A more explainable loan model may be a little less accurate; the framework does not decide that tradeoff for you.

The Core: four functions

The AI RMF Core has four functions, GOVERN, MAP, MEASURE and MANAGE, each broken down into categories and subcategories. NIST AI 100-1 (AI RMF 1.0), Executive Summary; Part 2, section 5 (AI RMF Core) NIST states that the actions under them are not a checklist and not necessarily an ordered set of steps. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5 GOVERN is cross cutting and runs through the other three; once its outcomes are in place, most users start with MAP and continue to MEASURE or MANAGE, and the process repeats over the whole life of the system. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5 Users may select among the categories and subcategories or apply all of them. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5

FunctionThe question it answers for Tessmoor
GOVERNWho decides, under which policies, and how do we know what AI we run?
MAPWhat is this system for, who does it affect, and what could go wrong?
MEASUREHow do we test it, and how will we know when it drifts?
MANAGEWhat do we do about each risk, and when do we stop?

The Playbook

NIST's AI RMF Playbook is an online companion that suggests tactical actions to reach the framework's outcomes; like the framework, it is voluntary, and organizations use its suggestions according to their needs. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 Treat a Playbook action as an idea to adapt, not a requirement to tick.

Where NIST sits next to the rules that bind

A voluntary framework sits under the rules that do bind. In Canada, privacy regulators have said that generative AI tools do not sit outside current legislative frameworks, and organizations using them must comply with applicable privacy laws. OPC and FPT privacy authorities, Principles for responsible, trustworthy and privacy-protective generative AI technologies (December 7, 2023) In the European Union, a high risk AI system needs a documented, continuous risk management system, and making sure it has one is a duty of its provider. AI Act Art. 9(1), (2), (5) and (8) AI Act Art. 16(a) to (l) For systems listed in Annex III, those rules apply from December 2, 2027. Regulation (EU) 2024/1689 Art. 2(2), Art. 111(4), Art. 113, as amended by Regulation (EU) 2026/1744 ISO/IEC 42001 is an international management system standard, and certification against it is performed by external certification bodies, not by ISO. ISO/IEC 42001:2023; ISO, Certification; ISO, Foreword, Supplementary information ISO/IEC 23894 gives guidance on AI risk management. ISO/IEC 23894:2023, iso.org catalogue page (Abstract; General information); ISO/IEC 42001:2023 catalogue page (FAQ) NIST work can feed all of these, but following NIST does not by itself show that any of them is met.

What "aligned with NIST" can honestly mean

Because the framework is voluntary and lets users select what they apply, "aligned with NIST" is a claim an organization makes about itself. NIST AI RMF 1.0 (NIST AI 100-1) NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5 An honest claim names the systems in scope, the functions and categories applied, and the records that show it. This lesson's template, the adoption brief, writes that down before the phrase goes into a contract or on a website.

Worked example

Tessmoor's adoption brief

Fictional organization. Tessmoor Credit Union is invented for this course. Any resemblance to a real organization is a coincidence.

The board's minute said "adopt the NIST framework". The AI risk lead turned it into one page before anyone repeated it outside the credit union.

QuestionTessmoor's answer
Framework and versionAI RMF 1.0 (NIST AI 100-1), voluntary NIST AI RMF 1.0 (NIST AI 100-1); revision watched at each quarterly review NIST AI RMF 1.0 (NIST AI 100-1)
Systems in scopeThe small business loan scoring model and the member assistant
What we apply nowAll of GOVERN; MAP, MEASURE and MANAGE for both systems, selected category by category NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5
Rules that come firstCanadian privacy law for both systems OPC and FPT privacy authorities, Principles for responsible, trustworthy and privacy-protective generative AI technologies (December 7, 2023); our lending policies and contracts
OwnerAI risk lead, reporting to the chief risk officer

What Tessmoor will say, and will not say. It will say "we use the NIST AI RMF 1.0 for our loan model and member assistant, and hold a profile for each". It will not say "NIST certified" or "compliant with NIST", because it can show only its own records, and the framework is voluntary. NIST AI RMF 1.0 (NIST AI 100-1)

Scenario exercise · Allow 7 min

Rewrite three claims

Three draft sentences reach the AI risk lead in one week:

  1. Marketing wants the website to say: "Our member assistant is NIST certified."
  2. A vendor of a fraud model writes: "Our model is compliant with the NIST AI RMF, so your privacy obligations are covered."
  3. The board pack says: "We will complete the NIST checklist in order, starting with GOVERN 1 and ending with MANAGE 4."

For each, say what is wrong and write a sentence Tessmoor could use instead.

Show the model answer
  1. "Certified" claims what Tessmoor cannot show. The framework is intended for voluntary use, and Tessmoor holds only its own records. NIST AI RMF 1.0 (NIST AI 100-1) Better: "We manage the risks of our member assistant using the NIST AI RMF 1.0, and keep a profile for it."
  2. A voluntary framework does not cover privacy obligations. Generative and other AI tools stay inside Canadian privacy law. OPC and FPT privacy authorities, Principles for responsible, trustworthy and privacy-protective generative AI technologies (December 7, 2023) Better: ask the vendor for its profile and test results, and check privacy law on its own terms.
  3. The Core is not a checklist in a fixed order. Actions are not a checklist or an ordered set of steps, and GOVERN runs through the other functions. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5 Better: "We will put the GOVERN outcomes in place, then map, measure and manage each system, and repeat."

Why this is the strongest answer

Every claim about NIST should survive the question "what record shows that?" Start from the framework's own status, voluntary and selectable, and the rules that bind, then describe only what you actually do. NIST AI RMF 1.0 (NIST AI 100-1) NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5

Template to download

AI RMF adoption brief

One page that says what your organization means when it says it uses the NIST AI RMF: version, systems in scope, what is applied now, the rules that come first, and the claims you will and will not make.

Check your understanding

Scenario questions. Answer them all, then check: you will see why each option is right or wrong.

Question 1 of 5

A customer questionnaire asks whether Tessmoor is "NIST certified". What is the accurate answer?

Question 2 of 5

The risk team wants to start the member assistant with MEASURE and map it afterwards. Does the framework allow it?

Question 3 of 5

A more explainable version of the loan model is slightly less accurate. What does the framework say about that choice?

Question 4 of 5

Who can feel the harm the AI RMF asks you to consider?

Question 5 of 5

Tessmoor's chief risk officer asks whether certification against ISO/IEC 42001 is something ISO itself performs. What is accurate?

Sources cited in this lesson

  1. NIST AI RMF 1.0 (NIST AI 100-1) checked on 2026-10-29
  2. NIST AI 100-1 (AI RMF 1.0), Executive Summary; Part 2, section 5 (AI RMF Core) checked on 2026-11-01
  3. NIST AI RMF 1.0 (NIST AI 100-1) checked on 2026-11-01
  4. NIST AI 100-1 (AI RMF 1.0), Part 1, section 1.1 checked on 2026-11-03
  5. NIST AI 100-1 (AI RMF 1.0), Part 1, section 3 and Figure 4 checked on 2026-11-03
  6. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 and Figure 5 checked on 2026-11-03
  7. NIST AI 100-1 (AI RMF 1.0), Part 2, section 5 checked on 2026-11-03
  8. OPC and FPT privacy authorities, Principles for responsible, trustworthy and privacy-protective generative AI technologies (December 7, 2023) official text current to 2025-05-06, checked on 2026-11-01
  9. AI Act Art. 9(1), (2), (5) and (8) official text current to 2026-07-27, checked on 2026-11-01
  10. AI Act Art. 16(a) to (l) official text current to 2026-07-27, checked on 2026-11-01
  11. Regulation (EU) 2024/1689 Art. 2(2), Art. 111(4), Art. 113, as amended by Regulation (EU) 2026/1744 official text current to 2026-07-27, checked on 2026-11-01
  12. ISO/IEC 42001:2023; ISO, Certification; ISO, Foreword, Supplementary information checked on 2026-11-01
  13. ISO/IEC 23894:2023, iso.org catalogue page (Abstract; General information); ISO/IEC 42001:2023 catalogue page (FAQ) checked on 2026-11-01

Get the full course

One payment of 59 USD. Every lesson, the knowledge checks, the final exam and the certificate open on your account. The first lesson is free to read.

Prices are in US dollars. Terms of sale · Privacy notice

You can ask for a full refund of a course within 14 days of buying it, if you have not passed its final exam.