NIST AI Risk Management Framework Core Functions: A Comprehensive Guide
Why NIST AI RMF Core Functions Matter
The National Institute of Standards and Technology (NIST) Artificial Intelligence Risk Management Framework (AI RMF) is one of the most significant voluntary frameworks for managing risks associated with AI systems. Published in January 2023, it provides organizations with a structured, flexible, and technology-agnostic approach to identifying, assessing, and mitigating AI-related risks. Understanding the core functions of the NIST AI RMF is essential for AI governance professionals because:
• It is a widely recognized standard referenced by policymakers, regulators, and organizations globally.
• It promotes trustworthy and responsible AI development and deployment.
• It helps organizations systematically manage AI risks throughout the AI lifecycle.
• It aligns with other frameworks and standards, making it a foundational knowledge area for AI governance certifications like the AIGP exam.
• It is voluntary and non-prescriptive, meaning organizations can adapt it to their unique contexts, which makes understanding its intent and structure critically important.
What Is the NIST AI RMF?
The NIST AI RMF (NIST AI 100-1) is a framework designed to help organizations manage risks to individuals, organizations, and society associated with AI. It is composed of two main parts:
1. Part 1: Foundational Information – Describes how organizations can frame AI risks and outlines the characteristics of trustworthy AI systems.
2. Part 2: The AI RMF Core – Provides the four core functions, categories, and subcategories that serve as actionable outcomes for AI risk management.
The framework emphasizes that AI risks are different from traditional software risks because AI systems can be opaque, complex, and context-dependent, and their behavior can change over time as they learn from new data.
Characteristics of Trustworthy AI (Key Background Knowledge)
Before diving into the core functions, it is important to understand that the NIST AI RMF identifies the following characteristics of trustworthy AI:
• Valid and Reliable – The AI system performs as intended and produces consistent results.
• Safe – The system does not endanger human life, health, property, or the environment.
• Secure and Resilient – The system can withstand adverse events and maintain functionality.
• Accountable and Transparent – Decisions and processes are explainable and traceable.
• Explainable and Interpretable – Outputs can be understood by relevant stakeholders.
• Privacy-Enhanced – The system respects privacy norms and protects personal data.
• Fair – with Harmful Bias Managed – The system avoids unjust discrimination and manages bias proactively.
These characteristics are interconnected and sometimes in tension with one another. The core functions help organizations operationalize these characteristics.
The Four Core Functions of the NIST AI RMF
The AI RMF Core is organized into four functions: GOVERN, MAP, MEASURE, and MANAGE. These functions are designed to be used together and are not sequential — they are iterative and can be applied throughout the AI lifecycle.
1. GOVERN
Purpose: Establish and maintain the organizational structures, policies, processes, and culture needed to manage AI risks effectively.
The GOVERN function is cross-cutting — it applies to and informs all other functions (MAP, MEASURE, and MANAGE). It is the foundational function that sets the tone for the entire AI risk management program.
Key Categories under GOVERN:
• GOVERN 1: Policies, processes, procedures, and practices are in place and regularly updated to govern AI risk management. This includes establishing clear accountability structures, roles, and responsibilities.
• GOVERN 2: Accountability structures are in place so that the appropriate teams and individuals are empowered, responsible, and trained for mapping, measuring, and managing AI risks.
• GOVERN 3: Workforce diversity, equity, inclusion, and accessibility processes are prioritized in AI risk management. This recognizes that diverse teams help identify and mitigate a broader range of risks.
• GOVERN 4: Organizational teams are committed to a culture that considers and communicates AI risk. This includes fostering an environment where raising concerns about AI risks is encouraged.
• GOVERN 5: Processes are in place for robust engagement with relevant AI actors. This includes stakeholder engagement, both internal and external, including affected communities.
• GOVERN 6: Policies and procedures are in place to address AI risks and benefits arising from third-party software and data. This covers supply chain and vendor management for AI components.
Key Takeaway: GOVERN is not just about having policies — it is about creating an organizational culture of responsible AI. It is the only function that is explicitly cross-cutting across the other three.
2. MAP
Purpose: Establish the context for framing AI risks by identifying and understanding the AI system, its intended purposes, its potential impacts, and the broader environment in which it operates.
The MAP function is about contextualizing risks. Before you can measure or manage risks, you must first understand what the AI system does, who it affects, and what could go wrong.
Key Categories under MAP:
• MAP 1: Context is established and understood. This includes understanding the intended purpose, the operational environment, stakeholders, legal and regulatory requirements, and the assumptions underpinning the AI system.
• MAP 2: Categorization of the AI system is performed. This involves classifying the AI system based on its risk level, complexity, and potential impact on individuals and communities.
• MAP 3: AI capabilities, targeted usage, goals, and expected benefits and costs compared with appropriate benchmarks are understood. This means conducting a thorough assessment of what the AI system is supposed to do versus what it might actually do.
• MAP 4: Risks and benefits are mapped for all components of the AI system including third-party software and data. This includes understanding interdependencies and potential cascading risks.
• MAP 5: Likelihood and magnitude of each identified risk are assessed. Impacts to individuals, groups, communities, organizations, and society are characterized.
Key Takeaway: MAP is fundamentally about understanding before acting. It emphasizes that risk management must be contextual — the same AI system could pose different risks in different deployment contexts.
3. MEASURE
Purpose: Employ quantitative, qualitative, or mixed-method tools, techniques, and methodologies to analyze, assess, benchmark, and monitor AI risk and related impacts.
The MEASURE function is about assessing and tracking risks using appropriate metrics and methodologies. It converts the contextual understanding from the MAP function into measurable and actionable risk information.
Key Categories under MEASURE:
• MEASURE 1: Appropriate methods and metrics are identified and applied to measure AI risks and trustworthiness characteristics. This includes selecting and validating metrics for fairness, accuracy, robustness, and other trustworthiness properties.
• MEASURE 2: AI systems are evaluated for trustworthy characteristics. This involves testing, auditing, and assessing AI systems against the identified metrics, including bias testing, performance evaluation, and security assessments.
• MEASURE 3: Mechanisms for tracking identified AI risks over time are in place. This emphasizes ongoing monitoring and not just one-time assessments, recognizing that AI risks can evolve as models drift or as deployment contexts change.
• MEASURE 4: Feedback about efficacy of measurement is collected and assessed. This means evaluating whether the metrics and measurement approaches themselves are adequate and adjusting them as needed.
Key Takeaway: MEASURE is not a one-time activity. It requires continuous monitoring and evaluation. The function also recognizes that measurement in AI is challenging — some risks may be difficult to quantify, and metrics may need to evolve over time.
4. MANAGE
Purpose: Allocate risk resources to mapped and measured risks on a regular basis and as defined by the GOVERN function. This involves planning, prioritizing, and implementing risk treatment actions.
The MANAGE function is about taking action on the risks identified through MAP and quantified through MEASURE. It involves prioritizing risks, implementing mitigation strategies, and communicating risk-related information to stakeholders.
Key Categories under MANAGE:
• MANAGE 1: AI risks based on assessments and other analytical output from the MAP and MEASURE functions are prioritized, responded to, and managed. This includes deciding whether to mitigate, transfer, accept, or avoid specific risks.
• MANAGE 2: Strategies to maximize AI benefits and minimize negative impacts are planned, prepared, implemented, documented, and informed by input from relevant AI actors. Risk treatments are documented and regularly monitored.
• MANAGE 3: AI risks and benefits from third-party resources are regularly monitored, and risk treatment is applied and documented. This extends risk management to the AI supply chain.
• MANAGE 4: Risk treatments, including response and recovery, and communication plans for the identified and measured AI risks are documented and monitored regularly. This includes incident response planning and stakeholder communication strategies.
Key Takeaway: MANAGE is about action and accountability. It requires that risk treatment decisions are documented, communicated, and revisited regularly. It also includes planning for when things go wrong (incident response and recovery).
How the Four Functions Work Together
The four functions are designed to work iteratively and concurrently, not as a linear sequence:
• GOVERN provides the overarching structure and culture that enables the other three functions. It is cross-cutting and persistent.
• MAP establishes the context — what the AI system does, who is affected, and what risks exist.
• MEASURE quantifies and monitors the risks identified during MAP.
• MANAGE takes action on the measured risks and ensures that responses are documented and tracked.
Think of it as: GOVERN sets the rules, MAP identifies the terrain, MEASURE takes readings, and MANAGE navigates the path forward. All four functions operate continuously and inform each other.
Important Distinctions to Remember
• The NIST AI RMF is voluntary, not mandatory (though it may be referenced by regulations).
• It is technology-agnostic — it applies to all types of AI, not just specific technologies like large language models.
• It is risk-based and outcome-oriented, not checklist-based.
• It is designed for all organizations that design, develop, deploy, or use AI systems, regardless of size or sector.
• The framework recognizes that not all AI risks can be eliminated — the goal is to manage them to acceptable levels.
• GOVERN is the only cross-cutting function — this is a frequently tested concept.
Companion Resources
NIST also provides several companion resources that support the AI RMF:
• NIST AI RMF Playbook – Provides suggested actions and references for each subcategory.
• AI RMF Crosswalks – Maps the AI RMF to other frameworks and standards (e.g., ISO/IEC 23894, OECD AI Principles).
• NIST AI RMF Profiles – Provide tailored implementation guidance for specific use cases or sectors (e.g., the Generative AI Profile, NIST AI 600-1).
Exam Tips: Answering Questions on NIST AI RMF Core Functions
Tip 1: Memorize the Four Functions and Their Purposes
Know that the four functions are GOVERN, MAP, MEASURE, and MANAGE. Be able to state the primary purpose of each in one sentence. A helpful mnemonic: "Good Managers Must Manage" (Govern, Map, Measure, Manage).
Tip 2: Remember That GOVERN Is Cross-Cutting
If a question asks which function is cross-cutting, foundational, or applies to all other functions, the answer is always GOVERN. This is one of the most commonly tested concepts. GOVERN informs and enables MAP, MEASURE, and MANAGE.
Tip 3: Distinguish Between MAP, MEASURE, and MANAGE
These three functions follow a logical progression: identify and contextualize risks (MAP) → assess and quantify risks (MEASURE) → take action on risks (MANAGE). If a question describes understanding context or identifying stakeholders, think MAP. If it describes testing, metrics, or monitoring, think MEASURE. If it describes prioritizing risks, implementing mitigations, or incident response, think MANAGE.
Tip 4: Know the Characteristics of Trustworthy AI
Questions may ask you to identify trustworthy AI characteristics or connect them to the core functions. Remember the seven characteristics: Valid and Reliable, Safe, Secure and Resilient, Accountable and Transparent, Explainable and Interpretable, Privacy-Enhanced, and Fair with Harmful Bias Managed.
Tip 5: Understand That the Framework Is Voluntary and Flexible
If a question suggests the NIST AI RMF is mandatory or prescriptive, that answer is likely wrong. The framework is voluntary, adaptable, and intended to complement existing organizational risk management processes.
Tip 6: Pay Attention to Third-Party and Supply Chain Risks
Both GOVERN (GOVERN 6) and MANAGE (MANAGE 3) explicitly address third-party and supply chain risks. Questions about vendor management, third-party AI components, or supply chain risk management should trigger consideration of these categories.
Tip 7: Recognize the Iterative Nature of the Framework
The functions are not strictly sequential. If a question implies that organizations must complete one function entirely before starting another, that answer is likely incorrect. The functions are designed to be iterative and applied concurrently.
Tip 8: Connect GOVERN to Culture and Accountability
GOVERN is not just about policies and procedures — it also encompasses organizational culture, diversity and inclusion (GOVERN 3), risk communication culture (GOVERN 4), and stakeholder engagement (GOVERN 5). Questions about organizational culture around AI risk management typically point to GOVERN.
Tip 9: Remember That MEASURE Includes Ongoing Monitoring
MEASURE is not just about initial testing. It explicitly includes mechanisms for tracking risks over time (MEASURE 3) and evaluating the effectiveness of measurements themselves (MEASURE 4). Questions about continuous monitoring or model drift assessment relate to MEASURE.
Tip 10: Watch for Distractor Answers That Mix Up Functions
Exam questions may include answer choices that describe activities belonging to a different function than the one asked about. For example, a question about MAP might include an answer choice that describes a MEASURE activity (like applying metrics). Always check whether the described activity matches the specific purpose of the function in question: context-setting (MAP), assessment (MEASURE), or action (MANAGE).
Tip 11: Know the Companion Resources
Be aware that the NIST AI RMF has companion resources like the Playbook and Profiles (including the Generative AI Profile, NIST AI 600-1). Questions may reference these as tools that support implementation of the core framework.
Tip 12: Practice Scenario-Based Reasoning
Many exam questions will present a scenario and ask which function or category applies. Practice by reading scenarios and asking yourself: Is this about setting up governance structures (GOVERN)? Understanding context and risks (MAP)? Assessing or measuring risks (MEASURE)? Taking action on risks (MANAGE)? This analytical approach will help you select the correct answer even for unfamiliar scenarios.
