NIST AI RMF 1.0 specifies what an organisation should achieve under each subcategory of the four functions. It does not specify how. The Playbook and profiles are the two NIST resources that fill that gap, in different ways and at different levels of specificity.
The AI RMF Playbook provides general implementation guidance for each subcategory of the framework: suggested actions, references, and documentation considerations. It is technology-neutral and use-case agnostic, mirroring the framework itself.
Profiles provide context-specific implementation guidance for particular use cases, sectors, or technologies. They identify risks distinctive to a context, prioritise relevant subcategories, and provide suggested actions tailored to the context. The Generative AI Profile (NIST AI 600-1) is the most prominent example.
The two resources are complementary. The Playbook is the default starting point for general AI RMF adoption; profiles are the natural extension when the organisation’s context matches a published profile.
“The AI RMF Playbook provides suggested tactical actions organizations can take to implement AI RMF subcategories. The Playbook is neither a checklist nor a set of steps to be followed in its entirety.” — NIST AI RMF Playbook, Introduction
Key definitions
| Term | Meaning |
|---|---|
| Playbook | NIST’s companion implementation guide to the AI RMF, providing suggested actions, references, and documentation considerations for each subcategory of the four functions. Published as a living document and updated as practice evolves. |
| Profile | An adaptation of the AI RMF to a specific context — a use case, sector, technology, or AI actor role. Profiles select, prioritise, and elaborate framework subcategories for the context. |
| Use case profile | A profile addressing a specific application or deployment context. |
| Target profile | A profile describing a desired state an organisation is working toward, often used for planning and gap analysis. |
| Suggested action | A concrete activity proposed in the Playbook or a profile for achieving a specific subcategory outcome. Suggested, not required. |
| Documentation consideration | Guidance on what should be recorded about implementation of a specific subcategory, supporting internal accountability and external attestation. |
| Transparency consideration | Guidance on what should be communicated externally about implementation of a specific subcategory. |
NIST AI RMF Playbook’s structure
The Playbook follows the framework’s structure exactly. For each subcategory of each function — Govern 1.1 through Manage 4.3 — the Playbook provides four elements:
| Element | What it offers |
|---|---|
| Suggested actions | Concrete activities the organisation may take to achieve the subcategory’s outcome. Range from foundational (developing policies) to operational (specific evaluation techniques). |
| AI actor tasks | Identification of which AI actor categories the suggested actions most directly involve. |
| Documentation considerations | What should be recorded about implementation — supporting internal accountability and external attestation. |
| Transparency considerations | What should be communicated externally about the implementation — to users, regulators, or other interested parties. |
Each subcategory’s Playbook content typically runs to several paragraphs across the four elements. Across the framework’s full set of subcategories — over 70 in total — the Playbook is a substantial document.
The Playbook is published and maintained by NIST at https://airc.nist.gov/airmf-resources/playbook/. It is treated as a living document and updated as practice evolves, as new resources become available, and as community input refines the suggested actions.
How the Playbook is used in practice
The Playbook serves four operational functions for organisations adopting the AI RMF.
Implementation starter for each subcategory. Organisations beginning AI RMF adoption typically use the Playbook to translate subcategory outcomes into operational activities without designing each activity from scratch. The suggested actions function as a starting menu — organisations select actions appropriate to their context, supplement with context-specific activities, and document their selections.
Reference library for AI risk management. The Playbook’s references — to NIST publications, academic literature, sector guidance, technical standards, and other frameworks — function as a curated entry point into the broader AI risk management literature. Organisations use the references to deepen specific areas (bias measurement, robustness testing, explainability methods) without surveying the field independently.
Documentation and transparency guidance. The documentation and transparency considerations specify what should be recorded internally and communicated externally for each subcategory. For organisations using AI RMF alignment as evidence to customers, partners, or regulators, these considerations are the most directly actionable part of the Playbook.
Adaptation reference for profiles. Profiles build on Playbook content, selecting and modifying suggested actions appropriate to the profile’s context. The Generative AI Profile follows this pattern, and sector-specific profiles produced by external organisations commonly do the same.
What NIST AI RMF Playbook does not provide
The Playbook is not a methodology. It does not prescribe how to perform bias measurement, how to conduct adversarial testing, how to construct an AI impact assessment, or how to design a specific evaluation procedure. It points to methods, references organisations that have developed methods, and identifies the considerations that method selection should address. Organisations seeking step-by-step methodology guidance need to combine the Playbook with sector-specific or technique-specific resources it references.
The Playbook is not a checklist. NIST is explicit that the suggested actions are not a sequence to be followed in order or a complete list to be implemented in full. Organisations select actions based on context, prioritise based on risk, and document their selections — applying the Playbook as a checklist over-implements and misses the contextual judgement the framework requires.
The Playbook is not a substitute for the framework itself. It elaborates the framework’s subcategories but does not replace them. Organisations adopting the Playbook without the framework lose the structural context that makes the Playbook’s content coherent.
NIST AI RMF Profiles: what they add
Profiles take the framework’s general structure and specify it for a particular context. The concept of profiles is established in AI RMF 1.0 itself, which describes profiles as a mechanism for tailoring the framework to specific applications, sectors, or technology types.
“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.” — NIST AI RMF 1.0, Section 5
A profile typically does four things:
| Function of a profile | What it produces |
|---|---|
| Identify risks distinctive to the context | A list of risks specific to the use case, sector, or technology — the substantive starting point for context-specific risk management |
| Select and prioritise subcategories | Identification of which subcategories bear most heavily on the context, which apply with modifications, which may be less applicable |
| Provide context-specific actions | Suggested actions tailored to the context, referencing techniques and considerations specific to the use case |
| Support comparison and benchmarking | A shared reference for organisations operating in the same context |
The Generative AI Profile (NIST AI 600-1)
NIST AI 600-1, published in July 2024, is the most developed published profile. It adapts the AI RMF to generative AI use cases and identifies twelve risks unique to or exacerbated by generative AI:
| Risk | Focus |
|---|---|
| CBRN information or capabilities | Lowered barriers to chemical, biological, radiological, nuclear weapons information |
| Confabulation | Confident generation of false content presented as fact |
| Dangerous, violent, or hateful content | Content facilitating violence, self-harm, or hate |
| Data privacy | Training data exposure, memorisation, output inference |
| Environmental impacts | Energy and resource consumption of training and inference |
| Harmful bias and homogenization | Output bias and reduction in viewpoint diversity |
| Human-AI configuration | Inappropriate reliance, anthropomorphisation, misaligned interaction |
| Information integrity | Erosion of the broader information ecosystem through synthetic content |
| Information security | Generative AI as attack target and attack tool |
| Intellectual property | Training data IP and output IP issues |
| Obscene, degrading, or abusive content | Non-consensual intimate imagery, CSAM, similar content |
| Value chain and component integration | Risks across the generative AI supply chain |
For each risk, the profile provides suggested actions mapped to the four functions. The profile is published at https://www.nist.gov/itl/airc and is referenced extensively in subsequent NIST and US federal AI policy work.
Other NIST AI RMF profiles
Beyond the Generative AI Profile, the profile ecosystem is still maturing. As of mid-2026, additional profiles exist or are in development through:
| Source | Examples |
|---|---|
| NIST | Ongoing work on profiles for specific sectors and technologies, tracked through the AI RMF Roadmap |
| Other US federal agencies | Agency-specific profiles addressing AI use within particular federal contexts |
| Sector regulators and industry bodies | Profiles for AI in healthcare, financial services, public safety, and other sectors |
| International organisations | Profiles produced by international bodies aligning the AI RMF with their own frameworks |
| External organisations | Profiles produced by consultancies, professional bodies, and industry groups |
Authoritative status varies. NIST-published profiles carry official status. Profiles from other sources are useful but should be evaluated for fit and rigour before adoption. The authoritative source for any specific AI governance instrument is the instrument itself; profiles are tools for navigation, not authoritative statements of equivalence.
How to combine the Playbook and profiles
A practical sequence for organisations adopting the AI RMF:
| Step | Resource | Purpose |
|---|---|---|
| 1. Determine relevant profile(s) | Profile catalogue | Identify whether a published profile matches the organisation’s context |
| 2. Apply the profile’s risk list as starting point | Profile | Use the profile’s identified risks as the basis for Map activities, supplemented by organisation-specific risks |
| 3. Apply the profile’s subcategory prioritisation | Profile | Focus implementation effort on the subcategories the profile identifies as most relevant |
| 4. Use the Playbook for general subcategory implementation | Playbook | For subcategories not specifically addressed by the profile, use Playbook content as the default starting point |
| 5. Combine profile actions with Playbook actions | Both | Profile suggested actions take precedence where both address the same subcategory; Playbook actions supplement |
| 6. Document selections and rationale | Profile and Playbook documentation considerations | Record which actions were selected, which were not, and why |
Where no published profile matches the organisation’s context, the Playbook serves as the default reference. Organisations operating in contexts without a profile commonly produce internal profiles describing their own adaptation, drawing on the Playbook and any partially relevant published profiles as starting points.
NIST AI RMF resource references
The primary references for working with the Playbook and profiles:
| Resource | Source |
|---|---|
| NIST AI RMF 1.0 | https://www.nist.gov/itl/airc — the core framework |
| AI RMF Playbook | https://airc.nist.gov/airmf-resources/playbook/ — companion implementation guidance |
| NIST AI 600-1: Generative AI Profile | https://www.nist.gov/itl/airc — generative AI use case profile |
| AI RMF Roadmap | https://www.nist.gov/itl/airc — planning document for framework evolution |
| AI RMF Crosswalks | https://www.nist.gov/itl/airc — mappings to ISO/IEC 42001, EU AI Act, OECD principles, others |
| AI Resource Center (AIRC) | https://airc.nist.gov — central hub for AI RMF resources and supporting materials |
NIST AI RMF resources are updated continuously. Organisations relying on specific Playbook actions, profile content, or crosswalk mappings should check current versions rather than treating an initial download as definitive.
FAQ
Is the Playbook mandatory if I adopt the AI RMF?
No. The Playbook is voluntary guidance about how to apply voluntary guidance. NIST is explicit that the Playbook is neither a checklist nor a set of steps to be followed in its entirety. Organisations select Playbook actions appropriate to their context and document their selections.
Do I have to use the Playbook to adopt the AI RMF?
No, but in practice most organisations use it. The framework itself specifies subcategory outcomes at a level of generality that allows broad applicability but provides limited implementation guidance. The Playbook fills that gap. Organisations adopting the framework without the Playbook either develop equivalent guidance internally or rely on external implementation references.
What is the relationship between the Playbook and ISO/IEC 42001 Annex A?
They serve different functions. Annex A is a normative list of 38 reference controls that must be considered for inclusion in an AIMS; the Playbook is voluntary implementation guidance for subcategory outcomes. Annex A produces a Statement of Applicability; the Playbook does not. Organisations using both frameworks typically treat Playbook content as methodology guidance for ISO 42001 Clauses 6 and 8 rather than as a substitute for Annex A.
When should I use a profile rather than just the Playbook?
Where a published profile matches the organisation’s context, the profile is more directly useful than the Playbook because it addresses the specific risks and prioritisations relevant to that context. Where no published profile matches, the Playbook is the default reference. Many organisations use both — the profile for context-specific guidance, the Playbook for subcategories the profile does not address in detail.
Are profiles certifiable?
No. The AI RMF and its profiles are voluntary throughout. Profile adoption is documented internally and through self-attestation or third-party assessment, not through certification.
Can I create my own profile?
Yes, you can. NIST describes profiles as adaptations developed by Framework users for their own contexts. Organisations operating in contexts without a published profile commonly produce internal profiles describing their adaptation of the framework. Internal profiles support consistency across an organisation’s AI systems and provide a basis for communication with customers and partners.
How often is the Playbook updated?
NIST treats the Playbook as a living document and updates it as practice evolves and as community input refines the suggested actions. There is no fixed update cycle. Organisations relying on specific Playbook actions should check the current version periodically rather than treating an initial download as definitive.
Does the Generative AI Profile replace the core AI RMF for generative AI use cases?
No, it doesn’t. The profile is an adaptation that depends on the core framework for its structure. Organisations developing or deploying generative AI implement the four functions of the AI RMF and use the Generative AI Profile to specify the implementation for the generative AI context. The profile is additive, not substitutive.
What if the Playbook and a profile give different guidance?
A profile’s guidance is more context-specific than the Playbook’s general guidance and takes precedence within the profile’s context. Where the profile addresses a subcategory, the profile’s suggested actions are the relevant reference for that context. Where the profile does not address a subcategory, the Playbook’s general guidance applies.
Are external profiles as authoritative as NIST-published profiles?
No, they are not. NIST-published profiles carry official status; profiles from other sources have varying authority depending on the source. For high-stakes work — regulatory submissions, customer due diligence, audit preparation — the authoritative source for AI governance is the underlying framework or regulation itself, not the profile. Profiles are tools for navigation and implementation, not authoritative statements of equivalence or compliance.
