The NIST AI Risk Management Framework in one read

The Artificial Intelligence Risk Management Framework (AI RMF) is a voluntary guide from the National Institute of Standards and Technology (NIST) for managing the risks of AI systems. Version 1.0 was published as NIST AI 100-1, dated January 2023, and NIST's framework page gives the release date as January 26, 2023. The framework organizes risk work into four functions: Govern, Map, Measure, and Manage. Two companions sit beside it: the online Playbook and the Generative AI Profile, NIST AI 600-1. As of October 21, 2026, the same NIST page says AI RMF 1.0 is being revised as part of the White House AI Action Plan.

What AI RMF 1.0 is

The framework states that the National Artificial Intelligence Initiative Act of 2020 directed its creation. It is meant to be voluntary, rights-preserving, non-sector-specific, and use-case agnostic. NIST's release announcement says NIST worked on it for 18 months and that it reflects about 400 sets of formal comments from more than 240 organizations.

The document has two parts. Part 1 frames AI risk and describes the audience. Part 2 holds the Core, meaning the four functions, and a section on profiles.

Part 1 lists seven characteristics of trustworthy AI systems:

• Valid and reliable
• Safe
• Secure and resilient
• Accountable and transparent
• Explainable and interpretable
• Privacy-enhanced
• Fair, with harmful bias managed

The framework says trustworthy AI requires balancing these characteristics against the system's context of use. The framework's diagram places valid and reliable at the base, as a necessary condition for the rest, and shows accountable and transparent as relating to all the others. The framework expected a review with formal input from the AI community no later than 2028.

Govern, Map, Measure, Manage

Each function breaks into categories and subcategories. The Core tables number six Govern categories, five Map, four Measure, and four Manage. The framework says the actions do not form a checklist or an ordered set of steps.

• Govern is a cross-cutting function, meant to inform and be infused throughout the other three.
• Map establishes the context to frame risks. When it is done, users should know enough about a system's impacts to make an initial go/no-go decision.
• Measure uses quantitative, qualitative, or mixed methods to analyze, assess, benchmark, and monitor risk. Systems should be tested before deployment and regularly in operation, with measures of uncertainty, comparison to benchmarks, and documented results. Independent review can improve the testing.
• Manage allocates risk resources to mapped and measured risks, including plans to respond to, recover from, and communicate about incidents.

The AI RMF Playbook supplies suggested actions for each subcategory. NIST says the Playbook is neither a checklist nor a set of steps to follow in full, and publishes it as PDF, CSV, Excel, and JSON. The page now says the Playbook will be updated after the framework is revised.

The framework also defines profiles. A use-case profile applies the Core to one setting. A current profile and a target profile describe present and desired practice, so comparing them shows the gaps. A cross-sectoral profile covers risks common across sectors, and NIST names the use of large language models as an example.

The Generative AI Profile

NIST AI 600-1, published July 26, 2024, is that kind of cross-sectoral profile. The profile calls itself a companion resource to AI RMF 1.0 and focuses on risks with an existing empirical evidence base. It lists 12 risks unique to or exacerbated by generative AI:

1. CBRN (chemical, biological, radiological, or nuclear) information or capabilities
2. Confabulation
3. Dangerous, violent, or hateful content
4. Data privacy
5. Environmental impacts
6. Harmful bias or homogenization
7. Human-AI configuration
8. Information integrity
9. Information security
10. Intellectual property
11. Obscene, degrading, and/or abusive content
12. Value chain and component integration

Each risk is mapped to the trustworthiness characteristics. The suggested actions carry identifiers tied to Core subcategories. Under MEASURE 2.5, which asks that a deployed system be shown to be valid and reliable, action MS-2.5-003 says to review and verify sources and citations in system outputs before deployment and during monitoring.

A revision under way

The profile says it was written pursuant to Executive Order (EO) 14110 of October 2023. Executive Order 14148, dated January 20, 2025, revoked EO 14110. NIST's publication record for the profile, updated April 8, 2026, still describes it as written pursuant to EO 14110.

In July 2025, America's AI Action Plan recommended that NIST revise the AI RMF to eliminate references to misinformation, Diversity, Equity, and Inclusion, and climate change. NIST's framework page now says version 1.0 is being revised under that plan. The same page reports an April 7, 2026 concept note for a new AI RMF profile on trustworthy AI in critical infrastructure.

The framework anticipated revision. It uses a two-number versioning system, in which the first number changes only with a major revision and minor revisions add a decimal, such as 1.1. A reader citing the framework after the revision appears should give the version number.

What to check

• Does a document cite AI RMF 1.0 by its designation, NIST AI 100-1, and say whether it follows the revised version once NIST publishes one?
• Which of the seven trustworthiness characteristics does a vendor claim, and what Measure evidence backs each one?
• Is there a current profile and a target profile, or only a statement of alignment?
• Does a generative AI system's risk register cover the 12 risks in NIST AI 600-1, including confabulation and information security?
• Were outputs' sources and citations checked before deployment and during monitoring, as action MS-2.5-003 suggests?

Azcuna's ELLMENT, the E-2D Large Language Model Entity, was proposed to NAVAIR PMA-231 under Navy SBIR Phase I topic DON26BZ01-NV010. The prototype, built at private expense on publicly available doctrine, answers with a page-level citation and refuses when the documents do not support an answer. The proposal has been submitted.

Sources

1. 1Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 · National Institute of Standards and Technology (January 2023)
2. 2AI Risk Management Framework · NIST Information Technology Laboratory
3. 3NIST Risk Management Framework Aims to Improve Trustworthiness of Artificial Intelligence · NIST news release (January 26, 2023)
4. 4NIST AI RMF Playbook · NIST Trustworthy and Responsible AI Resource Center
5. 5Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1 · National Institute of Standards and Technology (July 2024)
6. 6Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (publication record) · NIST (published July 26, 2024; page updated April 8, 2026)
7. 7Winning the Race: America's AI Action Plan · The White House (July 2025)
8. 8Executive Order 14148, Initial Rescissions of Harmful Executive Orders and Actions · Federal Register, 90 FR 8237 (January 28, 2025)

Where this shows up in our work

• CALIBER, the market testbed for AI agents measures from an AI agent's trades whether it is gullible, herds, overpays, or can be steered.
• ELLMENT, the offline LLM for E-2D crews answers a crew's question from the aircraft's own manuals and cites the page, with no network.