Model and data securityAdvancedLesson 1775 min read

AI risk frameworks: NIST AI RMF and ISO/IEC 42001

Two widely used frameworks help you organise AI risk work and prove it to others. Learn what they are and how security teams use them.

✦ Ask the tutor🧪 Lab · 30 min📝 Lesson test
Lesson in motion

In 60 seconds

AI risk frameworks: NIST AI RMF and ISO/IEC 42001

Two widely used frameworks help you organise AI risk work and prove it to others. Learn what they are and how security teams use them.

1/5
In simple words
A cricket team has a coach, a plan for each match, a way to score how it's going and a way to fix problems. Risk frameworks give a company the same kind of plan for its AI.
Frameworks do not make AI safe by themselves. They give you a shared structure, so everyone (engineers, security, legal, leadership and auditors) talks about risk the same way and nothing big is forgotten.

NIST AI Risk Management Framework

The NIST AI RMF 1.0 was published by the US National Institute of Standards and Technology in 2023. It is voluntary and used widely, including outside the US. It is organised into four functions.
FunctionMain ideaSecurity team examples
GovernCulture, roles, policies and accountability; applies across the other threeAI security policy, owners for each model, risk appetite, training
MapUnderstand the context: what the system is, who it affects, what could go wrongAI inventory, threat models, data flows, third-party dependencies
MeasureAnalyse, test and track risksRed teaming, evaluation results, robustness and privacy tests, metrics
ManagePrioritise and act on risks, and respond when things go wrongFixes, compensating controls, incident response, decommissioning
The framework also describes characteristics of trustworthy AI, including being valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed. Security teams own "secure and resilient" but contribute to the others.
In 2024 NIST published the Generative AI Profile (NIST AI 600-1). It applies the AI RMF to generative AI, describing risks specific to it and suggesting actions mapped to the four functions.

ISO/IEC 42001

ISO/IEC 42001 (2023) is an international standard for an AI management system. Like ISO/IEC 27001 for information security, it is a management system standard: it sets requirements for policies, roles, risk assessment, controls, monitoring and continual improvement. Organisations can be certified against it by an accredited certification body.
NIST AI RMF
  • Voluntary guidance
  • Flexible; no certification
  • Rich detail on AI risks and actions
  • Good for designing your risk work
ISO/IEC 42001
  • Formal requirements (shall statements)
  • Certifiable by third parties
  • Management system structure, similar to ISO 27001
  • Good for proving your AI governance to customers
Many organisations use both: NIST AI RMF to decide what to do, and ISO/IEC 42001 to run it as a managed, auditable system. If you already have ISO 27001, you can extend existing processes rather than start again.

MITRE ATLAS for threats

MITRE ATLAS is a knowledge base of adversary tactics and techniques against AI systems, modelled on MITRE ATT&CK, with case studies. Use it in the Map and Measure work: to build threat models, plan red team tests and check your detection coverage.
  1. 1

    Inventory

    List AI systems (Map).
  2. 2

    Threats

    Use ATLAS and OWASP lists to find relevant threats (Map).
  3. 3

    Test

    Red team and evaluate against them (Measure).
  4. 4

    Treat

    Fix, add controls or accept with sign-off (Manage).
  5. 5

    Oversee

    Report, review and improve (Govern).

Watch and read more

AI Risk Management FrameworkNIST · readingMITRE ATLASMITRE · knowledge base · reading

Lab

One real AI use case mapped to all four NIST AI RMF functions, with ATLAS-informed threats.

~30 min

The problem

Pick an AI use case, such as a loan document checker for a bank, or a support chatbot for an IT services company.

  1. Write two or three concrete activities under each function: Govern, Map, Measure, Manage.
  2. Browse MITRE ATLAS and pick three techniques relevant to your use case. Note where each is handled (Map for identifying, Measure for testing, Manage for treating).
  3. Name the owner of each activity.
  4. Mark which activities you already do, and which are gaps.

You are done when

Hard questions

Try to answer before you reveal. If you can answer these, you understood the lesson.

Q1Why is Govern described as applying across the other three functions?Reveal

Map, Measure and Manage only work if someone is responsible, has a policy to follow and resources to act. Govern sets those conditions: roles, accountability, culture and risk appetite. Without it, the other functions become one-off activities that nobody owns or repeats.

Q2A customer asks, "Are you NIST AI RMF certified?" How do you answer accurately?Reveal

Explain that the NIST AI RMF is voluntary guidance and has no certification. You can describe how you align with it and share evidence. If the customer wants a certification, ISO/IEC 42001 is a certifiable AI management system standard; say whether you hold it or plan to.

Please sign in to continue.

Questions people ask

Do we have to use NIST AI RMF if we are in India?

No. It is voluntary. But many global customers and partners know it, so it is a useful common language. Always check whether a regulator or contract asks for something specific.

Is ISO/IEC 42001 certification worth it?

It depends on your customers. If enterprise buyers or regulators ask for evidence of AI governance, certification can save repeated questionnaires. It takes real effort, so start by using it as a guide and decide on certification later.

How does this fit with our existing ISO 27001 work?

Well. Both are management system standards with a similar structure, so you can share processes such as risk assessment, internal audit and management review. Add the AI-specific parts: impact assessment, data governance for AI, and the AI system life cycle.

Is MITRE ATLAS a compliance framework?

No. It is a threat knowledge base. It helps you understand and test attacks. Frameworks like the AI RMF and ISO/IEC 42001 tell you how to manage risk; ATLAS helps you know what risks to look for.

Lesson test

5 questions. Get 3 right (60%) to pass and complete this lesson.

Sign in to take the test and save your progress