NIST AI Risk Management Framework Alignment
Applying the NIST AI Risk Management Framework to the systems you actually run, producing profiles and evidence that answer a customer's or a regulator's questions rather than a self-assessed maturity score.
The NIST framework is voluntary, which is why it is often deprioritised, and increasingly asked about in US enterprise procurement and federal-adjacent contracting, which is why it eventually gets done under time pressure. It is also, unusually for a governance framework, genuinely practical.
The NIST AI Risk Management Framework is a voluntary framework published by the US National Institute of Standards and Technology. It organises AI risk work into four functions — Govern, Map, Measure and Manage — and is applied through profiles that tailor it to a particular use case, sector or organisation.
The four functions, applied rather than recited
| Function | The question it answers | What we produce |
|---|---|---|
| Govern | Who is accountable and what is the culture and policy? | Accountability model, policy, escalation, competence |
| Map | What is this system, in what context, with what impacts? | Context and impact documentation per system |
| Measure | How do we know it works and where does it fail? | Metrics, test results, bias and robustness evidence |
| Manage | What do we do about the risks we found? | Treatment decisions, monitoring, incident response |
Govern runs across the other three rather than sitting first in a sequence. In practice Measure is where most organisations are weakest, because it demands evidence rather than description — and evidence is what a customer's questionnaire is actually asking for.
Profiles are what make it usable
A use-case profile, not a generic one
The framework is deliberately general. A profile tailors it to a specific context — this system, this sector, this risk appetite — and without one, alignment work produces documentation that could describe any organisation and reassures nobody.
Current and target profiles
Where you are and where you intend to be, with the gap sequenced by consequence. This is more useful than a maturity score because it produces a plan rather than a number, and because the target is chosen rather than assumed to be maximal.
Trustworthiness characteristics, evidenced
Validity and reliability, safety, security and resilience, accountability and transparency, explainability and interpretability, privacy, and fairness with harmful bias managed. Each needs measurement rather than assertion — see bias and fairness audit and explainable AI.
Generative AI needs its own treatment
NIST's generative AI profile addresses risks that the core framework handles only generally: confabulation, data leakage, harmful content, provenance and homogenisation. Where your AI is generative, this is the part that carries the real content.
Voluntary is not the same as optional in practice
The framework has no legal force, and it is increasingly cited in US enterprise procurement, federal-adjacent contracting and sector guidance. Organisations that align early answer these questions with evidence; organisations that wait answer them retrospectively, under deadline, with whatever documentation happens to exist.
How it sits alongside the other frameworks
- Against the EU AI Act. The Act is law with classification and conformity duties; NIST is a voluntary risk process. A NIST-aligned organisation has much of the substrate the Act's high-risk requirements assume, but not the legal obligations themselves.
- Against ISO/IEC 42001. 42001 is a certifiable management system; NIST is a risk framework with no certificate. They complement each other cleanly, and the evidence largely serves both.
- Against your existing enterprise risk process. NIST is designed to integrate rather than replace. Where you already run an operational risk framework, AI risk should extend it rather than living in a separate register nobody reads.
We build once against the union of what these ask for, so alignment work is not repeated for each. Doing them serially is a common and avoidable expense.
How the engagement runs
Profiles are built for real systems, because a generic profile reassures nobody.
Inventory and scoping
Systems catalogued and prioritised; the commercial or contractual driver established.
Current profile
Govern, Map, Measure and Manage assessed against reality rather than against intent.
Target profile
Chosen deliberately by consequence and appetite, with the gap sequenced.
Implementation
Gaps closed, with the Measure function prioritised because it is usually weakest and most asked about.
Evidence and handover
Documentation organised to answer customer and regulator questions; review cadence established.
What you receive
Current and target profiles with evidence behind the Measure function, not a maturity score.
Current profile
Honest assessment across all four functions for the systems in scope.
Target profile
Chosen by consequence and appetite, with the gap sequenced and owned.
Trustworthiness evidence
Measurement results for validity, safety, security, fairness, privacy and explainability.
Generative AI profile work
Confabulation, leakage, harmful content and provenance addressed where AI is generative.
Integration with enterprise risk
AI risk extending your existing register rather than sitting beside it.
Response pack
Documentation organised to answer procurement questionnaires and regulatory enquiries.
Is this the right engagement?
Worth being direct. NIST AI Risk Management Framework Alignment is the wrong spend in some situations, and those are listed rather than buried.
Good fit if
- US enterprise customers or federal-adjacent contracts ask about NIST alignment.
- A recognised, practical AI risk framework is wanted without pursuing certification.
- AI risk is currently assessed inconsistently or not at all.
- Generative AI is deployed and its specific risks have not been addressed.
- An existing enterprise risk process needs extending to cover AI.
Choose something else if
- A certificate is the commercial requirement. See ISO/IEC 42001.
- EU regulatory exposure is the driver. See EU AI Act compliance.
- You want a maturity score for a slide rather than evidence anyone will inspect.
- No AI system is in production and none is planned this year.
Frequently asked questions
Marked up with FAQPage schema so these answers can surface directly in search results and inside AI assistant responses.
What is the NIST AI Risk Management Framework?
A voluntary framework from the US National Institute of Standards and Technology that organises AI risk work into four functions — Govern, Map, Measure and Manage — applied through profiles that tailor it to a specific use case or organisation. It is practical rather than prescriptive, which is both its strength and the reason it needs tailoring to be useful.
Is NIST AI RMF mandatory?
No. It has no legal force. It is increasingly cited in US enterprise procurement, federal-adjacent contracting and sector guidance, which makes it commercially significant in practice. Organisations that align early answer these questions with evidence rather than retrospectively under a deadline.
How does it compare with ISO 42001?
ISO 42001 is a certifiable management system standard — you get an auditable certificate. NIST is a risk framework with no certification path. They complement each other and the underlying evidence largely serves both, so organisations needing both should build once rather than run two programmes.
Does it cover generative AI?
The core framework covers it generally, and NIST's generative AI profile addresses the specific risks — confabulation, data leakage, harmful content, provenance and homogenisation. Where your AI is generative, that profile carries most of the substance and is where we concentrate the work.
Where do organisations usually fall short?
In the Measure function. Govern and Map produce documentation, which organisations are good at; Measure demands actual evidence that a system performs as claimed and fails in known ways. That evidence is precisely what a customer questionnaire or a regulator asks for, so it is where we prioritise effort.
Often paired with this
Most clients combine two or three engagements from the AI Governance, Security and Compliance pillar. These are the ones that most often run immediately before or after.
AI Governance Framework Design
Proportionate risk tiers, named accountability and decision rights — a framework people follow rather than route around.
Read more →AI Risk Assessment and Model Cards
Risk assessments that end in decisions, and model cards that state limits honestly rather than advertising.
Read more →Responsible AI and Bias and Fairness Audit
The fairness definition chosen explicitly, disparities measured on real subgroups, remediation tested not asserted.
Read more →Is this the right engagement?
Tell us what you are trying to build. If a different service fits better, or if you do not need us at all, we will say so.