The NIST AI Risk Management Framework and its Generative AI Profile, explained

NIST’s voluntary AI RMF 1.0 (January 2023) and its Generative AI Profile, NIST AI 600-1 (July 2024), are still the published versions as of November 14, 2026, while NIST revises the framework under the 2025 AI Action Plan and drafts profiles for critical infrastructure and cybersecurity.

NIST released the AI Risk Management Framework, AI RMF 1.0, on January 26, 2023, and released a companion Generative AI Profile, NIST AI 600-1, on July 26, 2024 [1]. NIST says the framework is intended for voluntary use [1]. As checked on November 14, 2026, NIST’s framework page says AI RMF 1.0 is being revised as part of the White House AI Action Plan [1]. This explainer covers three parts: what the framework itself asks of an organization, what the generative AI profile adds, and what is changing around both, with the status of each piece on the date we checked.

What AI RMF 1.0 asks of an organization

The framework was developed as directed by the National Artificial Intelligence Initiative Act of 2020 (P.L. 116-283), to offer a resource to organizations designing, developing, deploying or using AI systems [2]. NIST describes it as voluntary, rights-preserving, non-sector-specific and use-case agnostic, meant for organizations of all sizes and in all sectors [2]. Nothing in it is a legal requirement on its own; a regulator, contract or internal policy may make it one for a given organization.

It has two parts. Part 1 frames AI risk and describes seven characteristics of trustworthy AI systems: valid and reliable; safe; secure and resilient; accountable and transparent; explainable and interpretable; privacy-enhanced; and fair with harmful bias managed [2]. NIST treats valid and reliable as a necessary condition for the others, and accountable and transparent as relating to all of them [2].

Part 2 is the Core: four functions, Govern, Map, Measure and Manage, each broken into categories and subcategories [2]. Govern is designed to be cross-cutting and to inform the other three, which can be applied to a specific AI system at a specific stage of its lifecycle [2]. NIST says the actions in the Core are not a checklist and not necessarily an ordered set of steps; after putting the Govern outcomes in place, most users would start with Map and continue to Measure or Manage [2]. Organizations may select among the categories and subcategories according to their resources and capabilities [2].

The framework also describes profiles, which is how it is meant to be tailored. Use-case profiles apply the functions to a specific setting or application; temporal profiles describe a current state and a target state, so that comparing the two shows the gaps to close; and cross-sectoral profiles cover risks of models or applications used across sectors [2]. A companion AI RMF Playbook suggests tactical actions for each outcome and is voluntary in the same way [2].

The governance outcomes reach procurement as well as development. The Govern function addresses the full product lifecycle, including legal and other issues concerning third-party software, hardware and data [2]. For a bank, health system or agency that buys most of its AI rather than building it, that is where the framework touches daily work.

What the Generative AI Profile adds

NIST AI 600-1 describes itself as a cross-sectoral profile of, and companion resource for, AI RMF 1.0 for generative AI, issued pursuant to Executive Order 14110 [3] [4]. It defines twelve risks that are unique to or made worse by generative AI [3]:

  • CBRN information or capabilities;
  • confabulation, the confident production of false content;
  • dangerous, violent or hateful content;
  • data privacy;
  • environmental impacts;
  • harmful bias or homogenization;
  • human-AI configuration, such as over-reliance or automation bias;
  • information integrity;
  • information security;
  • intellectual property;
  • obscene, degrading or abusive content; and
  • value chain and component integration, including improper supplier vetting [3].

Each risk is mapped to the trustworthy AI characteristics of the framework, and the profile then sets out suggested actions, organized by AI RMF subcategory and listed for only some subcategories [3]. It says the activities in AI RMF 1.0 and its Playbook already apply to generative AI risks [3]. The working group behind it limited its focus to four considerations, governance, content provenance, pre-deployment testing and incident disclosure, so the suggested actions concentrate there [3]. The profile says future revisions will add subcategories, risks and actions as evidence develops [3].

The AI Action Plan, published in July 2025, says President Trump has already rescinded Executive Order 14110, the order the profile names as its basis [3] [5]. NIST’s publication page for AI 600-1 still lists the July 26, 2024 profile, and its framework page still links it, as of November 14, 2026 [1] [4]. We found no NIST notice withdrawing it.

What is changing, and the status of each piece

The framework revision. Among its recommended policy actions, the AI Action Plan says the Department of Commerce, through NIST, should revise the AI RMF to eliminate references to misinformation, diversity, equity and inclusion, and climate change [5]. NIST’s framework page says AI RMF 1.0 is being revised as part of the plan [1]. As checked on November 14, 2026, that page offers AI RMF 1.0 for download and links no revised draft [1]. We have not seen revised text, so this page does not say what a new version will contain.

The recommendation and the current profile overlap. In our reading, three of the profile’s twelve risks deal with the subjects the plan names: information integrity, which covers dis- and misinformation campaigns; environmental impacts; and harmful bias [3] [5]. In our view, organizations that built control sets from AI 600-1 should watch those sections most closely when revised text appears, while on the same reading the other risks, such as confabulation, data privacy, information security and value chain integration, do not touch those subjects.

A critical infrastructure profile. NIST’s framework page says that on April 7, 2026 it released a concept note for an AI RMF Profile on Trustworthy AI in Critical Infrastructure, to guide operators toward risk management practices when using AI-enabled capabilities [1]. The project page itself shows a created date of April 6, 2026 and an updated date of July 17, 2026, and lists the project as ongoing [6]; the two NIST pages give different April dates, and we report both. The project page describes the profile as covering information technology, operational technology and industrial control systems, and invites operators to join a community of interest that will receive discussion drafts [6]. As checked, no draft profile is posted on that page.

A cybersecurity profile for AI. Separately, NIST published a preliminary draft of NIST IR 8596, the Cybersecurity Framework Profile for Artificial Intelligence, on December 16, 2025, with comments due January 30, 2026 [7]. It is organized on the Cybersecurity Framework 2.0 outcomes and has three focus areas: securing AI system components, conducting AI-enabled cyber defense, and thwarting AI-enabled cyberattacks [7]. NIST says comments on the preliminary draft will inform an initial public draft; as checked on November 14, 2026, the publication page lists only the preliminary draft [7].

Using the framework while it is being revised

In our view, the revision is a reason to structure AI risk work so it can absorb a new version, rather than a reason to wait. The four functions and the idea of current and target profiles are the framework’s core structure, and the Action Plan’s recommendation does not mention them [2] [5]. Mapping internal controls to Core subcategories by identifier, rather than copying NIST’s wording into policies, will make a later update easier to absorb.

Three practical points follow. First, where a regulator, contract or customer cites the AI RMF, record which version and which profile it cites. Second, security teams that already use the Cybersecurity Framework can follow the Cyber AI Profile draft, because it reuses that framework’s structure [7]. Third, operators in critical infrastructure sectors can join NIST’s community of interest if they want to see the critical infrastructure profile in draft [6].

Status as checked on November 14, 2026. This explainer describes NIST guidance and federal policy documents; it is not legal advice, and how any of it applies to a given organization is a question for its own counsel and risk staff.

Moraph’s teams help organizations with work like this. See our solutions.

Spotted an error? Tell us and we will check it and correct the explainer where needed.

  • NIST AI RMF
  • NIST AI 100-1
  • NIST AI 600-1
  • generative AI
  • AI Action Plan
  • NIST IR 8596
  • AI risk management