The NIST AI RMF as the AIGP tests it

The NIST AI Risk Management Framework is one of three named sources in the AIGP's smallest competency, and the one that most candidates can describe in a sentence and fail to answer a question on. The Body of Knowledge asks you to "Understand the NIST AI Risk Management Framework and Playbook (e.g., core functions, categories and subcategories)", which is a request to know the framework's structure, not its history. This guide sets out that structure from NIST AI 100-1, the framework's own text: what the framework is and is not, the seven characteristics of trustworthy AI it is built around, the four functions of its Core and the categories under each, the Playbook, and the profiles, with the examples the text itself gives.

The short answer

• Voluntary. NIST AI 100-1 calls itself "voluntary, rights-preserving, non-sector-specific, and use-case agnostic": no certification, no audit, no penalty.
• Seven characteristics of trustworthy AI, with valid and reliable as the base of the others.
• Four functions, GOVERN, MAP, MEASURE and MANAGE, with 19 categories under them; GOVERN cuts across the other three.
• The Playbook is the online, voluntary companion of suggested actions; profiles apply the Core to one setting or one point in time.
• The exam asks you to name the function behind a described activity, place a characteristic, and tell the framework from a law.

What the framework is

The framework is a US government document with no force of law. Its text says so in one sentence: "The Framework is intended to be voluntary, rights-preserving, non-sector-specific, and use-case agnostic, providing flexibility to organizations of all sizes and in all sectors and throughout society to implement the approaches in the Framework." It was written "As directed by the National Artificial Intelligence Initiative Act of 2020 (P.L. 116-283)", and its goal "is to offer a resource to the organizations designing, developing, deploying, or using AI systems to help manage the many risks of AI and promote trustworthy and responsible development and use of AI systems". The people it is written for are "AI actors", the framework's term for those who play a role in an AI system's life.

Three words in that description decide exam questions. Voluntary: the framework carries no certification, no audit and no penalty, which is the difference between it and a law such as the EU AI Act. Non-sector-specific: it is written for any organization. Use-case agnostic: it does not say which uses of AI are allowed, only how to manage the risk of whatever use an organization chooses.

The framework is still version 1.0, but NIST has said it will change. Its page for the framework, which gives the release date as January 26, 2023, now adds: "The AI RMF 1.0 is being revised as part of the White House AI Action Plan." No revised version had been published when this guide was updated, so AI 100-1 remains the text quoted here.

The seven characteristics of trustworthy AI

Before the Core, the framework defines what it is aiming at. "Characteristics of trustworthy AI systems include: valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed." The text's figure places "Valid & Reliable" as "a necessary condition" of the others, and each characteristic has its own section, 3.1 to 3.7, with definitions drawn where possible from international standards.

The exam uses this list two ways: as a list, asking which characteristic a described property belongs to, and as the content of the MEASURE function, which evaluates a system against them. Learn the pairs as the framework pairs them: valid and reliable, secure and resilient, accountable and transparent, explainable and interpretable; a question that splits a pair is testing whether you know they are one characteristic.

The Core: four functions

The framework's operating part is the Core, "composed of four functions: GOVERN, MAP, MEASURE, and MANAGE", each broken into categories, and each category into subcategories. The Core is where the exam's questions on the framework live, usually as a description of an activity with the function to name.

GOVERN is the foundation. The text lists what the function does; among its bullets, it "cultivates and implements a culture of risk management within organizations designing, developing, deploying, evaluating, or acquiring AI systems", it "outlines processes, documents, and organizational schemes that anticipate, identify, and manage the risks a system can pose", it "provides a structure by which AI risk management functions can align with organizational principles, policies, and strategic priorities", and it "addresses full product lifecycle and associated processes, including legal and other issues concerning use of third-party software or hardware systems and data". The sentence to carry into the exam is this one: "GOVERN is a cross-cutting function that is infused throughout AI risk management". The other three functions operate inside the structure GOVERN sets up. Its six categories run from legal and regulatory requirements, through accountability structures, workforce diversity, equity, inclusion and accessibility, an organizational culture of critical thinking and safety, "robust engagement with relevant AI actors", to policies for third-party software, data and other resources.

MAP "establishes the context to frame risks related to an AI system." The text explains why a separate function is needed: the AI lifecycle "consists of many interdependent activities involving a diverse set of actors", and "AI actors in charge of one part of the process often do not have full visibility or control over other parts and their associated contexts." MAP's five categories start with the plainest sentence in the framework, "Context is established and understood", and go on to categorizing the system, its "capabilities, targeted usage, goals, and expected benefits", the risks and benefits "for all components of the AI system", and the impacts to individuals, groups, communities, organizations and society.

MEASURE "employs quantitative, qualitative, or mixed-method tools, techniques, and methodologies to analyze, assess, benchmark, and monitor AI risk and related impacts." It "uses knowledge relevant to AI risks identified in the MAP function and informs the MANAGE function", and the text sets the cadence: "AI systems should be tested before their deployment and regularly while in operation." Its four categories cover the approaches and metrics for measurement, the evaluation of systems for the trustworthy characteristics, "Mechanisms for tracking identified AI risks over time", and feedback about whether the measurement itself works.

MANAGE "entails allocating risk resources to mapped and measured risks", on a schedule the GOVERN function defines. Its four categories cover prioritizing and responding to the risks that MAP and MEASURE produced, strategies to maximize benefits and minimize negative impacts, risks and benefits from third-party resources, and "Risk treatments, including response and recovery", with communication plans.

That gives 19 categories across the four functions, six, five, four and four, each with subcategories beneath it; the Body of Knowledge names "core functions, categories and subcategories" as the level it tests. You do not need the subcategory numbers. You need to recognize an activity and place it: setting policy, roles and culture is GOVERN; understanding the system's purpose, users and context is MAP; testing, metrics and tracking are MEASURE; deciding what to do about a risk, and doing it, is MANAGE.

The Playbook

The Body of Knowledge names the Playbook beside the framework, and the exam can ask what it is. AI 100-1 describes it: "An online companion resource to the AI RMF, the NIST AI RMF Playbook, is available to help organizations navigate the AI RMF and achieve its outcomes through suggested tactical actions they can apply within their own contexts. Like the AI RMF, the Playbook is voluntary and organizations can utilize the suggestions according to their needs and interests." It follows the Core's structure: for each function, the framework says its practices "are described in the NIST AI RMF Playbook". Playbook users "can create tailored guidance selected from suggested material for their own use and contribute their suggestions for sharing with the broader community", and the Playbook sits with the framework in "the NIST Trustworthy and Responsible AI Resource Center".

So the pair is framework and companion: AI 100-1 gives the outcomes, the Playbook gives suggested actions toward them, and neither is mandatory.

Profiles

The framework's last structural idea is the profile. "AI RMF use-case profiles are implementations of the AI RMF functions, categories, and subcategories for a specific setting or application based on the requirements, risk tolerance, and resources of the Framework user: for example, an AI RMF hiring profile or an AI RMF fair housing profile." A second kind, temporal profiles, "are descriptions of either the current state or the desired, target state of specific AI risk management activities", which lets an organization compare where it is with where it wants to be. A question that describes an organization tailoring the Core to one application, or comparing its current practice with a target, is describing a profile.

NIST has since published a profile of its own: NIST AI 600-1, the Generative Artificial Intelligence Profile, released on July 26, 2024, which its page says "can help organizations identify unique risks posed by generative AI". The same page lists an April 7, 2026 concept note for a profile on trustworthy AI in critical infrastructure.

How it is tested

The Body of Knowledge's indicator begins "Understand", which on the Bloom's taxonomy page of the same document means "Explain ideas or concepts" rather than apply them. Competency II.D carries 3 to 5 scored questions across its three indicators, so the framework may draw one or two. They tend to take four shapes:

1. Name the function. An activity is described; the options are the four functions. The verb that matches wins: governing, mapping, measuring or managing.
2. Place the characteristic. A property is described, say a system that keeps working under attack; the answer is the characteristic, here secure and resilient.
3. Distinguish the framework from a law or a standard. The distractors say the framework is mandatory, certifiable or sector-specific; the text says it is "voluntary, rights-preserving, non-sector-specific, and use-case agnostic".
4. Name the companion. The Playbook is the online, voluntary companion of suggested actions; a profile is the Core applied to one setting or one point in time.

The OECD's principles and the ISO standards are the other two sources in II.D, and the Body of Knowledge lists the three as a set; The OECD AI Principles as the AIGP tests it covers the first. Our course teaches the framework in its own lesson, with the categories and practice questions in the exam's format; the first competency is free in full with a free account, and The AIGP Body of Knowledge v2.1, explained shows where II.D sits in the blueprint.