Law and framework guide
NIST AI Risk Management Framework
Written with AI from the sources below and checked against our facts sheet. Last checked . How this site is made · Spot an error?
The NIST AI Risk Management Framework (AI RMF 1.0, NIST AI 100-1) is a voluntary framework for managing risks from AI systems. It comes from the US National Institute of Standards and Technology, and any organization in any sector can use it. Released on 26 January 2023, it is built around four functions: Govern, Map, Measure and Manage. NIST says a revision is underway, but version 1.0 remains current.
In short
- Voluntary, sector-agnostic and not certifiable, though some state laws give it weight.
- Four functions: GOVERN (cross-cutting), MAP, MEASURE and MANAGE, split into 19 categories and 72 subcategories.
- Seven characteristics of trustworthy AI, with trade-offs that each organization balances in context.
- Companion resources: the Playbook (suggested actions) and profiles, including the Generative AI Profile (NIST AI 600-1, July 2024).
| Publisher | US National Institute of Standards and Technology (NIST) |
|---|---|
| Document | NIST AI 100-1, AI RMF 1.0 |
| Released | 26 January 2023 |
| Status (Nov 2026) | Version 1.0 is current; NIST says it is being revised under the White House AI Action Plan |
| Legal status | Voluntary; no certification |
| Structure | 4 functions, 19 categories, 72 subcategories |
| Generative AI companion | NIST AI 600-1, Generative AI Profile, released 26 July 2024 |
What is the NIST AI RMF?
The NIST AI Risk Management Framework is a voluntary guide for managing the risks AI systems pose to people, organizations and society. NIST released version 1.0 on 26 January 2023 as NIST AI 100-1.
It is sector-agnostic and use-case agnostic. A bank, a hospital and a software start-up can all use it. NIST treats risk as a combination of an event's probability and the size of its consequences.
The framework has two parts. Part 1 explains how to frame AI risk and what trustworthy AI looks like. Part 2 is the Core: four functions broken into categories and subcategories, plus guidance on profiles.
NIST names harms to three groups: people (rights, safety, opportunity), organizations (operations, reputation) and ecosystems (supply chains, financial systems, the environment). Each organization sets its own risk tolerance; the framework does not prescribe one. Our lesson NIST AI RMF: the big picture covers these foundations.
Is the NIST AI RMF mandatory?
No. NIST says the AI RMF is intended for voluntary use. No US law requires private companies to adopt it, and NIST does not certify anyone against it.
It still carries legal weight in places:
- Texas. TRAIGA has been in effect since 1 January 2026. It shields a defendant that finds a violation through internal review while substantially complying with a recognized framework. The law names NIST's Generative AI Profile (Bus. & Com. Code § 552.105(e)).
- Colorado. The original AI Act, SB 24-205, named the AI RMF and ISO/IEC 42001 in its affirmative defense. Its replacement, SB 26-189, dropped that defense. See the Colorado AI Act hub.
As of 7 November 2026, version 1.0 is still current. NIST states that it "is being revised as part of the White House AI Action Plan," but no new version has been published.
What are the seven characteristics of trustworthy AI?
NIST lists seven characteristics a trustworthy AI system should show.
| Characteristic | What it means |
|---|---|
| Valid and reliable | Does what it is meant to do, consistently. The base for the others. |
| Safe | Does not endanger life, health, property or the environment. |
| Secure and resilient | Withstands attacks and recovers from unexpected events. |
| Accountable and transparent | Someone answers for it, and information about it is available. Cuts across the others. |
| Explainable and interpretable | People can understand how an output was produced and what it means for them. |
| Privacy-enhanced | Protects people's autonomy, identity and dignity through privacy safeguards. |
| Fair, with harmful bias managed | Addresses equality and equity, including systemic, computational and statistical, and human-cognitive bias. |
The characteristics involve trade-offs. More interpretability can cost some accuracy, and privacy techniques can reduce the data available for fairness testing. NIST does not rank them; organizations balance them in context. For one of them in depth, see explainable AI.
What are the four functions of the NIST AI RMF?
The Core has four functions. GOVERN applies across the whole life cycle. MAP, MEASURE and MANAGE can be used in any order and repeated as needed.
| Function | What it does | Categories | Example outcome |
|---|---|---|---|
| GOVERN | Builds the policies, accountability, culture and processes the rest rely on | 6 | Roles, training and risk tolerance are set; third-party risk policies exist |
| MAP | Establishes context and identifies risks and impacts | 5 | Intended purpose, users, setting and likely impacts are documented |
| MEASURE | Analyzes, tests and tracks risks with quantitative and qualitative methods | 4 | The system is evaluated for fairness, safety and privacy, and risks are tracked over time |
| MANAGE | Prioritizes risks and acts on them | 4 | Risks are mitigated, transferred, avoided or accepted; incidents are handled and communicated |
A memory aid from our lessons: "Good Maps Make Moves." Third-party risk shows up three times: GOVERN 6, MAP 4 and MANAGE 3. For the wider discipline, see AI risk management.
What are categories, the Playbook and profiles?
Categories and subcategories. The four functions break into 19 categories and 72 subcategories. Each is written as an outcome, not a step. For example, MANAGE 2.4 says mechanisms exist to supersede, disengage or deactivate systems that perform inconsistently with their intended use. Organizations pick the outcomes that fit their context.
The Playbook. The AI RMF Playbook is NIST's online companion. For each subcategory it suggests actions, documentation prompts and references. It is voluntary and not a checklist. The framework says what to aim for; the Playbook suggests how.
Profiles. A profile applies the framework to a specific setting. Use-case profiles target one application, such as hiring. Temporal profiles compare a current state with a target state to reveal gaps. Cross-sectoral profiles cover risks shared across sectors.
The Generative AI Profile. NIST AI 600-1, released on 26 July 2024, is a cross-sectoral profile for generative AI. It names 12 risks unique to or worsened by generative AI. Examples include confabulation, information integrity, CBRN information, human-AI configuration, and value chain and component integration. It maps suggested actions to AI RMF subcategories.
NIST has also released a concept note for a profile on trustworthy AI in critical infrastructure (7 April 2026). Our lesson Inside the RMF goes deeper on all of this.
How does the NIST AI RMF relate to ISO/IEC 42001 and the EU AI Act?
The three do different jobs.
| NIST AI RMF | ISO/IEC 42001 | EU AI Act | |
|---|---|---|---|
| What it is | Voluntary risk framework | International management-system standard | Binding EU regulation |
| Issued by | NIST (US government) | ISO and IEC | EU legislature |
| Certifiable? | No | Yes, by accredited certification bodies | High-risk systems need conformity assessment and CE marking |
| Focus | Outcomes for managing AI risk | An auditable system for governing AI across the organization | Legal duties by risk tier and role |
They can work together. An organization can use the AI RMF to structure its risk work and ISO/IEC 42001 to certify its management system. Where the EU AI Act applies, it sets the legal floor. NIST's resource center lists crosswalks that map the AI RMF to ISO/IEC 42001, ISO/IEC 42005 and other frameworks. NIST notes that a listing is not an endorsement.
Following the AI RMF does not by itself show EU AI Act compliance. See NIST AI RMF vs ISO/IEC 42001, the ISO/IEC 42001 hub and the EU AI Act hub.
How does the AIGP exam test the NIST AI RMF?
- GOVERN is cross-cutting. An option that treats it as a one-time first step is a trap.
- No tiers, no certificate. Implementation tiers belong to the NIST Cybersecurity Framework, not the AI RMF. Nobody is "AI RMF certified."
- Know the seven characteristics. Distractors like "profitable" or "fully autonomous" are never on the list.
- What, how, why. Transparency answers what happened; explainability, how the output was produced; interpretability, why it matters to the person.
- Framework, Playbook, profile. Outcomes, suggested actions, and tailoring to a setting, in that order.
Change log
- 7 Nov 2026: Page published.
Pages in this guide
Frequently asked questions
Is the NIST AI RMF mandatory?
No. NIST designed it for voluntary use, and no federal law requires private companies to adopt it. Some state laws give it weight. Texas's TRAIGA offers a defense to companies that substantially comply with NIST's Generative AI Profile or another recognized framework. Customers and contracts may also ask for alignment with it.
Can you get certified in the NIST AI RMF?
No. NIST does not certify organizations, systems or people against the AI RMF, and the framework has no implementation tiers or maturity scores. Organizations that want third-party certification of their AI governance usually look to ISO/IEC 42001, which accredited certification bodies audit.
What are the four functions of the NIST AI RMF?
GOVERN, MAP, MEASURE and MANAGE. GOVERN sets policies, accountability and culture, and applies throughout. MAP establishes context and identifies risks. MEASURE analyzes, tests and tracks them. MANAGE prioritizes and responds to them. Together the functions contain 19 categories and 72 subcategories.
What is NIST AI 600-1?
NIST AI 600-1 is the Generative AI Profile of the AI RMF, released on 26 July 2024. It names 12 risks unique to or worsened by generative AI, such as confabulation and information integrity. It maps suggested actions to AI RMF subcategories. It extends the framework rather than replacing it.
Is there a NIST AI RMF 2.0?
Not as of 7 November 2026. NIST lists AI RMF 1.0 (January 2023) as current. It states that the framework is being revised as part of the White House AI Action Plan. Check the NIST AI RMF page for updates.
Related
Sources
- NIST: AI Risk Management Framework
- NIST AI 100-1: Artificial Intelligence Risk Management Framework (AI RMF 1.0)
- NIST AI RMF Playbook (Trustworthy and Responsible AI Resource Center)
- NIST AI 600-1: Generative Artificial Intelligence Profile (Jul 2024)
- NIST AIRC: AI RMF crosswalks
- Texas Legislature: HB 149 (Texas Responsible AI Governance Act), enrolled text
- Colorado General Assembly: SB 24-205, Consumer Protections for Artificial Intelligence