NIST AI RMF Playbook — What It Is, How to Use It

Last verified: 2026-07-29 — primary source: NIST AI Resource Center (airc.nist.gov), retrieved 2026-07-29.

Quick answer: as of July 29, 2026, NIST distributes the AI RMF Playbook in four downloadable formats — PDF, Excel, CSV and JSON — alongside the browsable web version, and releases updates approximately twice per year.

The NIST AI Risk Management Framework Playbook is the U.S. National Institute of Standards and Technology's official companion resource to NIST AI RMF 1.0. The Playbook does not replace the framework — it operationalizes it, providing concrete suggested actions, references, and informative documentation pointers for every subcategory of every function.

NIST publishes and maintains the Playbook on the AI Resource Center (AIRC) at airc.nist.gov/AI_RMF_Knowledge_Base/Playbook (retrieved 2026-07-29). It is voluntary, free of charge, and maintained as a living resource separate from the underlying framework PDF (NIST AI 100-1) — the framework publication is version-locked, while the Playbook is re-released on its own roughly twice-yearly cadence.

What the Playbook actually contains

The AI RMF itself (NIST AI 100-1, January 26, 2023) defines four core functions — GOVERN, MAP, MEASURE, MANAGE — and a structured set of categories and subcategories underneath each function. The framework PDF tells you what outcomes to achieve. The Playbook tells you how to approach them.

For each subcategory, the Playbook supplies:

• Suggested actions — practical, role-relevant steps an organization might take to satisfy that subcategory's outcome. Suggestions are written as prompts for the team rather than mandatory controls.
• Transparency and documentation prompts — questions the organization should be able to answer about the system at this stage of the lifecycle.
• References — pointers to academic, standards-body, and industry sources that informed NIST's thinking. References include ISO/IEC standards, IEEE guidance, peer-reviewed literature, and federal-agency outputs.

The Playbook is not a checklist. It is a structured menu of options. NIST is explicit that organizations should adopt only the suggestions that fit their risk profile, sector, and operational reality.

What the Playbook is not

Three common misconceptions are worth correcting up front:

1. It is not certifiable. There is no "NIST AI RMF certification" issued by NIST itself. NIST does not certify, accredit, or audit conformance with the AI RMF. Organizations that need a certifiable management-system standard should use ISO/IEC 42001:2023, which is certifiable through accredited bodies.
2. It is not a replacement for sector regulation. Following the Playbook does not exempt an organization from the Colorado AI Act, Texas TRAIGA, NYC Local Law 144, HIPAA, FCRA, EEOC, or any other binding regime. The Playbook is voluntary guidance that supports compliance with those regimes — it does not substitute for them.
3. It is not static. NIST updates the Playbook on the AIRC website without re-issuing the AI RMF PDF, and states that Playbook updates are released approximately twice per year (airc.nist.gov, retrieved 2026-07-29). Organizations relying on a snapshot should re-check the AIRC on that cadence rather than annually, and should date every export they file as evidence.

NIST AI RMF certification — the honest answer

The term "NIST AI RMF certification" appears in several forms in the market. Compliance teams should know the distinction (see the NIST AI RMF detail page for the source-cited matrix that maps each market phrase to what is actually being attested):

• Issued by NIST itself: none. NIST does not run a certification program for the AI RMF.
• Issued by third-party training providers: several commercial providers offer courses and credentials such as "NIST AI RMF 1.0 Architect." These are private certifications attesting that an individual has completed the provider's training; they are not endorsed or audited by NIST. They are useful as professional development; they are not a substitute for organizational conformance.
• For the organization: pursue ISO/IEC 42001 certification through an accredited body, optionally with a NIST AI RMF crosswalk in the documentation.

Generative AI Profile — the GenAI extension

NIST released the Generative AI Profile (NIST AI 600-1) on July 26, 2024, as a profile of the AI RMF tailored for generative AI risks. The Profile keeps the four-function structure but adds GenAI-specific subcategory guidance for confabulation (hallucination), data privacy, dangerous or violent recommendations, environmental harms, harmful bias, human-AI configuration, information integrity, information security, intellectual property, obscene/degrading/abusive content, toxicity, and value chain and component integration risks.

The Playbook on AIRC has been progressively extended with GenAI-Profile-specific suggested actions following the 600-1 release.

How to use the Playbook in practice

For an organization standing up an AI risk program for the first time, a defensible workflow is:

1. Read AI RMF 1.0 (NIST AI 100-1) end-to-end. The framework PDF is the authoritative source for what each function and subcategory means. The Playbook assumes you already understand the framework structure.
2. Choose your scope. Decide which AI systems are in scope (start with consumer-affecting and high-impact systems first — these are also the systems most likely to fall under the Colorado AI Act high-risk definition).
3. Walk through the GOVERN function first. Most organizations underinvest in governance. GOVERN subcategories address policy, accountability, oversight, and culture — the foundation that makes MAP/MEASURE/MANAGE meaningful. Use the AI governance guide to translate that function into committee ownership, lifecycle gates, and escalation paths.
4. For each in-scope subcategory, review the Playbook's suggested actions. Pick the ones that fit your maturity, sector, and risk appetite. Document why others were rejected. The audit trail is itself a control.
5. Cross-reference with state-law obligations. For each adopted action, note which state-law requirement it supports. Examples: Playbook MEASURE 2.11 actions on bias testing → satisfy NYC Local Law 144 bias audit underlying methodology; MAP 5.1 actions on impact assessment → support Colorado AI Act § 6-1-1703(3) annual impact assessment; GOVERN actions on accountability and policy → preposition deployer duties for likely 2027 reintroduction of Virginia HB 2094-style bills and provide a defensible due-care baseline in jurisdictions that regulate AI through narrow topic-specific statutes rather than a comprehensive framework, such as Florida's HB 919 + HB 757 deepfake regime.
6. Refresh on NIST's cadence, not the calendar's. The Playbook is living guidance released roughly twice per year, and the GenAI Profile content is added incrementally. Re-download the Excel or CSV export each cycle and diff it against the archived copy — that diff, not an annual re-read, is what shows an auditor which suggested actions changed and when.

How the Playbook compares to alternatives

For most U.S.-only programs, NIST AI RMF + Playbook is the operational anchor. For multi-region compliance (EU, UK, multi-state) where a certificate is useful, layer ISO/IEC 42001 on top — the two are complementary, not competing.

Where to get the Playbook PDF (and the machine-readable versions)

NIST distributes the Playbook in four downloadable formats from the AI Resource Center, in addition to the browsable web version (all links retrieved 2026-07-29):

The Excel and CSV exports are the practical starting point for building a control library, because each suggested action arrives already keyed to its subcategory — the join key this Atlas uses when mapping Playbook actions to state-law duties.

Pair whichever export is archived with the version-locked publications behind it: the AI RMF 1.0 PDF (https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf) and the Generative AI Profile PDF (https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf). Because the Playbook is refreshed roughly twice a year while those two PDFs are static, a dated Playbook export plus the two framework PDFs is the reference set that reconstructs what guidance was in force on a given date.

Frequently asked questions

What is the NIST AI RMF Playbook?

The Playbook is NIST's official companion resource to AI RMF 1.0 that supplies suggested actions, transparency-and-documentation prompts, and references for each subcategory of the framework's four functions (GOVERN, MAP, MEASURE, MANAGE). It is voluntary, free, and hosted on the NIST AI Resource Center.

Where is the NIST AI RMF Playbook PDF?

NIST publishes it at https://airc.nist.gov/docs/AI_RMF_Playbook.pdf, free and without registration (retrieved 2026-07-29). The same AI Resource Center page also offers Excel (playbook.xlsx), CSV (playbook.csv) and JSON (playbook.json) exports of the identical suggested actions, plus the browsable web version. Because NIST re-releases the Playbook roughly twice per year, teams filing an export as audit evidence should date it and keep the superseded copy; the companion framework publications, NIST.AI.100-1.pdf and NIST.AI.600-1.pdf, are version-locked and do not change between Playbook releases.

Is there a NIST AI RMF certification?

Not from NIST. NIST does not certify organizations or products against the AI RMF. Some private training providers issue "NIST AI RMF 1.0 Architect"-style credentials for individuals — these are useful as professional development but are not NIST-endorsed and do not certify organizational conformance. For organizational certification, ISO/IEC 42001 is the certifiable AI management-system standard.

How does the Playbook differ from AI RMF 1.0 itself?

AI RMF 1.0 (NIST AI 100-1, January 26, 2023) defines the four functions, categories, and subcategory outcomes — the what. The Playbook supplies suggested actions and references for each subcategory — the how. The framework is stable and version-locked; the Playbook is living guidance.

Does adopting the Playbook satisfy the Colorado AI Act?

Not automatically. The Playbook is voluntary guidance and Colorado AI Act § 6-1-1703 imposes binding obligations on deployers of high-risk AI. However, Playbook suggested actions for MAP and MEASURE substantially overlap with Colorado's annual impact-assessment contents. Adopting Playbook actions and documenting which ones address each Colorado obligation is a defensible compliance approach — but the assessment, disclosure, and notification duties under § 6-1-1703 are still the binding floor.

Does Playbook adoption help with NYC Local Law 144 bias audits?

Playbook MEASURE-function suggested actions on bias evaluation align with the methodology that NYC LL 144 audits assume — but LL 144 specifically requires an independent bias audit by a third party, with public summary and candidate notice. Internal Playbook-driven testing does not substitute. See the NYC Local Law 144 detail for the audit and disclosure requirements.

What's the relationship between the Playbook and the GenAI Profile?

The GenAI Profile (NIST AI 600-1, July 26, 2024) is itself a profile of the AI RMF tailored to generative-AI risks. The Playbook's web resource has been extended with GenAI-Profile-specific suggested actions. For generative-AI deployments, both are required reading — the Profile defines GenAI-specific subcategory outcomes and the Playbook supplies the corresponding suggested actions.

Cross-references

• NIST AI RMF 1.0 detail — framework summary, controls, and crosswalk to state laws
• ISO/IEC 42001 — the certifiable alternative
• Federal vs state AI law — how voluntary frameworks interact with binding state laws
• AI impact assessment template — practical template that draws on Playbook MAP actions
• AI compliance framework — register structure for translating Playbook actions into evidence owners and refresh dates
• AI governance — operating model for GOVERN decision rights, committees, and lifecycle gates
• Colorado AI Act — § 6-1-1703(3) impact-assessment regime that Playbook actions help satisfy