What is the NIST AI Risk Management Framework?
DigitalXForce wrote this page.
The NIST AI Risk Management Framework, AI RMF 1.0, is NIST’s voluntary guidance on managing AI risk, released on 26 January 2023. Its core has 4 functions: Govern, Map, Measure and Manage. NIST added a generative AI profile in July 2024 and is revising the framework. DigitalXForce inventories AI systems and tests their controls, and the risk decisions stay yours.
An organization cannot govern AI it has not found. The AI RMF gives a shared way to describe the risk, and the practical work starts with an inventory and a set of controls that can be tested.
What the NIST AI RMF asks for
NIST released the AI RMF for voluntary use, to help organizations build trustworthiness into the design, development, use and evaluation of AI products, services and systems. Its Core has 4 functions, each broken into categories and subcategories. NIST notes that the actions “do not constitute a checklist”. After Govern is in place, most users start with Map and continue to Measure or Manage.
Govern
Govern runs through the other 3 functions. It covers policies and processes, accountability, workforce diversity, a culture that considers AI risk, engagement with the people affected, and risk from third-party software and data. Two of its subcategories are concrete. Govern 1.1 asks that legal and regulatory requirements involving AI are understood, managed and documented, and Govern 1.6 asks that “mechanisms are in place to inventory AI systems”.
Map
Map sets the context for each AI system. It covers the system’s purpose, its categorization, its capabilities, its expected benefits and costs, the risks of its components, including third-party software and data, and its impacts on individuals, groups, organizations and society.
Measure
Measure uses quantitative, qualitative or mixed methods to analyze, assess, benchmark and monitor AI risk. It evaluates systems for trustworthy characteristics and puts mechanisms in place to track identified risks over time.
Manage
Manage allocates resources to the risks that Map and Measure found, on a regular basis. It prioritizes and responds to those risks, manages the risks and benefits that come from third parties, and documents and monitors each risk treatment, including response and recovery.
Profiles and the next version
NIST released the Generative Artificial Intelligence Profile, NIST AI 600-1, on 26 July 2024, to help organizations identify risks that generative AI poses and plan actions to manage them. NIST says AI RMF 1.0 is being revised as part of the White House AI Action Plan. On 7 April 2026 it released a concept note for a profile on trustworthy AI in critical infrastructure.
Should you adopt the NIST AI RMF or ISO/IEC 42001 first?
Start with the NIST AI RMF when you need a shared way to find and describe AI risk and nobody is asking for a certificate. Start with ISO/IEC 42001 when customers or regulators want an AI management system that an accredited body certifies. The two fit together: the AI RMF’s 4 functions help decide what the ISO/IEC 42001 management system has to control.
| Comparison | NIST AI RMF 1.0 | ISO/IEC 42001:2023 |
|---|---|---|
| What it is | It is voluntary guidance from NIST for managing AI risk. | It is an international standard for an AI management system. |
| Who checks it | No one certifies against it. | Accredited certification bodies audit and certify against it. |
| How it is organized | It has 4 functions: Govern, Map, Measure and Manage. | It follows the clause structure of other ISO management system standards, with controls in an annex. |
| What you get | You get a shared vocabulary and a set of outcomes to aim for. | You get a certificate that customers can check. |
| Where it helps first | It helps a team build an AI inventory and a risk method quickly. | It helps a supplier answer customers who ask for proof. |
What can internal audit test over AI models?
Internal audit can test whether each AI system does what it was approved to do, stays within the limits set for it, and is watched after release. The Measure and Manage functions name what to look for, and each row gives the evidence an auditor can ask for.
| What to test | AI RMF subcategory | Evidence to ask for |
|---|---|---|
| The system is in the inventory, with an owner. | GOVERN 1.6 asks for mechanisms to inventory AI systems. | Ask for the inventory entry, the owner and the approval record. |
| The system was valid and reliable before release. | MEASURE 2.5 asks that the system to be deployed is shown to be valid and reliable. | Ask for the validation tests, the data used and the documented limits. |
| The model is explained and validated. | MEASURE 2.9 asks that the model is explained, validated and documented. | Ask for the model documentation and the validation sign-off. |
| Fairness and bias were evaluated. | MEASURE 2.11 asks that fairness and bias are evaluated and the results documented. | Ask for the bias test results and the decision made on them. |
| Security and resilience were evaluated. | MEASURE 2.7 asks that security and resilience are evaluated and documented. | Ask for the security test results, including tests for misuse of prompts. |
| Privacy risk was examined. | MEASURE 2.10 asks that privacy risk is examined and documented. | Ask for the privacy assessment and the data flows. |
| Behavior is watched in production, so drift is seen. | MEASURE 2.4 asks that functionality and behavior are monitored in production. | Ask for the monitoring records and the alerts raised in the period. |
| People can override, stop or retire the system. | MANAGE 4.1 asks for post-deployment monitoring plans that cover override, decommissioning and incident response. | Ask for the plan and the record of its last test. |
NIST’s Generative AI Profile, NIST AI 600-1, adds actions for generative systems under the same subcategory numbers.
How can NIST AI RMF controls be continuously assessed?
NIST AI RMF controls are continuously assessed by keeping the AI inventory current, watching each deployed system for drift, bias and safety problems, and testing each AI control against evidence, so a failure becomes a risk with an owner.
Continuous Control Assurance (CCA) uses evidence, monitoring and validation to determine whether controls continue to operate as expected. Continuous Control Monitoring (CCM) monitors conditions, evidence and signals associated with controls. CCM is a capability within CCA. For AI systems, CCM watches the signals from each deployed system, and CCA decides whether the controls around it still work.
The framework asks for this after release as well as before it. MEASURE 2.4 asks that functionality and behavior are monitored in production, and MANAGE 4.1 asks for post-deployment monitoring plans that cover override, decommissioning and incident response.
| AI RMF subcategory | What is checked | Evidence DigitalXForce reads |
|---|---|---|
| GOVERN 1.6 | Every AI system is in the inventory. | AI TRiSCM discovers AI assets across cloud, code repositories, CI/CD pipelines, containers, model endpoints, RAG and vector databases and business applications. |
| MEASURE 2.4 | Functionality and behavior are watched in production. | The module runs drift detection and performance and safety telemetry after deployment. |
| MEASURE 2.11 | Fairness and bias are evaluated. | The module runs bias monitoring after deployment. |
| MEASURE 2.7 | Security and resilience are evaluated. | The module assesses each system against security, safety, privacy, robustness and operational risk. |
| MEASURE 2.10 | Privacy risk is examined. | The same assessment covers privacy risk. |
| MANAGE 4.1 | Problems found after release are acted on. | A failed guardrail raises a risk in the enterprise risk register, and X-ROC alerts, triages, escalates and reports it. |
A monitoring signal does not decide what impact a system has on people or how much AI risk the organization accepts. Those are Govern and Map decisions, and they stay with the people accountable for each system. Each AI control has its own test frequency, set by how quickly its evidence can change.
What DigitalXForce does for the NIST AI RMF
AI governance is the set of policies, roles, controls and evidence an organization uses to decide which AI systems it runs, how they may be used and whether their controls keep working. DigitalXForce governs AI through its AI TRiSCM™ module (Trust, Risk, Security and Compliance Management). The module discovers AI assets across cloud, code repositories, CI/CD pipelines, containers, model endpoints, RAG and vector databases and business applications. That inventory is the one Govern 1.6 asks for.
The module assesses LLMs, chatbots, copilots, RAG systems, AI agents and model files against security, safety, privacy, robustness and operational risk. After deployment, the module keeps watching each system through bias monitoring, drift detection and performance and safety telemetry. Each AI control is mapped once to the NIST AI RMF, including its generative AI profile, and to ISO/IEC 42001, the EU AI Act, the OWASP Top 10 for LLM Applications and MITRE ATLAS.
An AI system is approved before it goes into production. An exception or a risk acceptance is recorded as a named person’s decision with a reason and an expiry date, and remediation is tracked against the AI risk it addresses.
AI systems enter the same chain as every other asset. A failed guardrail raises a risk in the enterprise risk register, AI JedAI scores it, and X-ROC, the XForce Risk Operations Center, alerts, triages, escalates and reports it. The vendors that supply models are assessed in the third-party risk management module, and a material change in a vendor’s posture triggers a reassessment. The AI sub-posture of the Digital Trust Score covers the AI inventory, model and use-case risk, data sensitivity exposure and human oversight.
XForce GPT writes the narrative on what AI is in use, what its risk is and what changed, and an analyst reviews AI output before anyone relies on it.
What stays with your organization
The AI RMF is written for the people who design, deploy and oversee AI, and most of its outcomes are decisions.
- Your executive leadership takes responsibility for decisions about the risks of developing and deploying AI, as Govern 2.3 describes.
- Your organization sets accountability, trains its people, builds diverse teams and keeps a culture that considers AI risk.
- Your teams engage with the people an AI system affects and fold their feedback into its design.
- Your teams set each system’s context and characterize its impacts on individuals, groups and society.
- Your teams choose the methods and metrics for measuring AI risk and decide whether a system is deployed, changed or retired.
How the NIST AI RMF maps to the platform
| AI RMF function | What DigitalXForce does | What the organization still does |
|---|---|---|
| Govern | AI TRiSCM discovers and inventories AI systems, maps AI controls to the regulations that apply, and requires approval before production. Model vendors are assessed in the TPRM module. | Leadership sets policy, risk tolerance and accountability, and it engages with the people affected. |
| Map | Each AI system is assessed against security, safety, privacy, robustness and operational risk, and third-party components are assessed through TPRM. | Teams set each system’s purpose and context and characterize its impacts. |
| Measure | Bias monitoring, drift detection and performance and safety telemetry track each system over time, and each test result is kept with its evidence. | Teams choose the methods and metrics and review whether the measurement works. |
| Manage | Failed guardrails become risks in the register, AI JedAI scores them, X-ROC alerts, triages, escalates and reports them, and exceptions carry a named owner and an expiry. | Teams decide each treatment and whether a system is deployed, changed or retired. |
Where DigitalXForce is not the answer
An organization with a handful of AI tools, no regulatory exposure and no board question about AI does not need a governance module. A written policy and an inventory in a spreadsheet will do for now. DigitalXForce fits an organization whose AI is spreading faster than its register, that reports under the EU AI Act, ISO/IEC 42001 or a sector regulator, or whose board wants the answer as evidence.
DigitalXForce also does not decide what a system’s impact on people is, or how much AI risk the organization will accept. Those are Govern and Map decisions, and they stay with people.
Questions about the NIST AI RMF
How can NIST AI RMF controls be continuously assessed?
Keep the AI inventory current, watch each deployed system for drift, bias and safety problems, and test each AI control against evidence. MEASURE 2.4 asks that functionality and behavior are monitored in production. In DigitalXForce, a failed guardrail raises a risk that AI JedAI scores, and X-ROC alerts, triages, escalates and reports it.
How do organizations map the NIST AI RMF to enterprise controls?
Write each subcategory that applies as a control with an owner, the evidence that proves it and the system that holds that evidence. GOVERN 1.6, for example, becomes a control that every AI system is in the inventory. DigitalXForce maps each AI control once to the NIST AI RMF, ISO/IEC 42001 and the EU AI Act, so one result serves all three.
Is the NIST AI RMF mandatory?
It is not. NIST released it for voluntary use, to help organizations build trustworthiness into AI products, services and systems. Each organization applies the functions in the way that suits its needs and resources.
What are the 4 functions of the NIST AI RMF?
They are Govern, Map, Measure and Manage. Govern is cross-cutting and runs through the other 3. Map sets the context, Measure analyzes and tracks the risk, and Manage prioritizes and treats it.
Does the NIST AI RMF cover generative AI?
It does, through the Generative Artificial Intelligence Profile, NIST AI 600-1, which NIST released on 26 July 2024. The profile describes risks specific to generative AI and proposes actions to manage them.
Is NIST changing the AI RMF?
NIST says AI RMF 1.0 is being revised as part of the White House AI Action Plan. It released a concept note for a profile on trustworthy AI in critical infrastructure on 7 April 2026.
How does the NIST AI RMF relate to the EU AI Act?
The AI RMF is voluntary US guidance, and the EU AI Act is binding EU law. Many of their concerns overlap, including inventory, risk assessment, human oversight and monitoring. DigitalXForce maps each AI control once to both, so one test result serves each.
DigitalXForce terms used on this page
DigitalXForce gives each of its terms one definition and uses it everywhere, including in its glossary.
- TRiSCM, or Trust, Risk, Security and Compliance Management, is a category of enterprise software defined by DigitalXForce that converges automated governance, risk and compliance with security posture management into a single real-time system, replacing periodic control validation with continuous control assurance.
- A risk operations center is an operating model for continuously measuring, prioritizing and reducing risk, in the way a security operations center handles threats; DigitalXForce’s implementation is X-ROC, the XForce Risk Operations Center. X-ROC is the operations layer of the DigitalXForce TRiSCM platform, where control failures, risk changes and vendor events are alerted, triaged, escalated and reported.
- AI JedAI is the DigitalXForce AI engine that analyzes: it reasons over control evidence and live telemetry, maps documents to controls and frameworks, scores and prioritizes risk, and recommends remediation mapped to framework requirements.
- XForce GPT is the DigitalXForce generative AI engine that writes: it produces the plain-language risk narratives and board-ready reports, generates policies, standards and plans, and runs the embedded assistant.
- The Digital Trust Score is DigitalXForce’s composite score from 300 to 850, computed continuously from live control evidence across seven sub-postures: security, compliance, audit, resilience, third-party, AI and risk.
Related pages
- The AI TRiSCM module is where DigitalXForce governs AI systems.
- AI governance and compliance and automated AI assessment describe the two main uses.
- What Is Continuous Control Assurance (CCA)? explains the strategic level of control assurance, which validates whether controls continue to operate as expected.
- What Is Continuous Control Monitoring (CCM)? explains the capability within Continuous Control Assurance that monitors the evidence behind each control between tests.
- The enterprise risk management module holds AI risks beside every other risk.
- X-ROC, the XForce Risk Operations Center, is where failed guardrails are worked.
- What is a common control framework? explains how one control, written and tested once, serves every framework it maps to.
- How internal audit and Continuous Control Assurance (CCA) work together explains when internal audit can rely on continuous assurance results.
- The NIST CSF 2.0 page covers NIST’s framework for cybersecurity outcomes.
- The EU AI Act page covers the binding rules that sit beside the AI RMF.
- The frameworks page lists the most requested frameworks.
Sources
- NIST’s AI Risk Management Framework page gives the release date of 26 January 2023, the voluntary status, the generative AI profile of 26 July 2024, the revision under the White House AI Action Plan and the concept note of 7 April 2026, at https://www.nist.gov/itl/ai-risk-management-framework.
- Section 5 of AI RMF 1.0, the Core, with the 4 functions and their categories and subcategories, is published in NIST’s Trustworthy and Responsible AI Resource Center at https://airc.nist.gov/airmf-resources/airmf/5-sec-core/.
- NIST AI 100-1, Artificial Intelligence Risk Management Framework (AI RMF 1.0), January 2023, is at https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf. GOVERN 1.6, MEASURE 2.4, 2.5, 2.7, 2.9, 2.10 and 2.11 and MANAGE 4.1 were read in the published PDF.
- NIST AI 600-1, the Generative Artificial Intelligence Profile, July 2024, is at https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf.
- ISO describes ISO/IEC 42001:2023, Information technology, Artificial intelligence, Management system, at https://www.iso.org/standard/42001. iso.org refused scripted requests on 25 September 2026, so this page was not read that day.
- The NIST sources were read on 25 September 2026.
Do this once instead of every audit.
Every step above can be done by hand. DigitalXForce does them continuously, maps the result to 50+ frameworks once, and keeps the evidence current between audits. A 30 minute walkthrough on your own framework set shows what that removes from your calendar.



