What does NIST AI RMF 1.0 say about AI in banking?

What are the key points of NIST AI RMF 1.0?

• Released January 26, 2023 after two public drafts (March and August 2022) and a 2021 request for information; developed under the National Artificial Intelligence Initiative Act of 2020.
• Part 1 explains how organizations should frame AI risk and lists seven trustworthiness characteristics; Part 2 is the Core, built from four functions: Govern (cross-cutting), Map, Measure and Manage.
• Each function breaks into categories and subcategories of outcomes (e.g., Govern 1.1: legal and regulatory requirements are understood, managed and documented) that organizations tailor to their context.
• Explicitly rights-preserving and sector-agnostic; intended to be used alongside existing frameworks such as NIST's Cybersecurity Framework and Privacy Framework.
• Companion resources: the online AI RMF Playbook, a Roadmap, and crosswalks to ISO/IEC standards and other frameworks, hosted at the NIST AI Resource Center (airc.nist.gov).
• 'Profiles' — use-case or sector implementations of the Core — are the mechanism for later work such as the Generative AI Profile (AI 600-1) and the 2026 critical-infrastructure profile.

What did NIST AI RMF 1.0 change for banks?

Before 2023, US banks had model risk management guidance (SR 11-7 / OCC 2011-12) but no widely accepted, examiner-legible structure for governing AI as an enterprise risk rather than as individual quantitative models. The AI RMF supplied that structure, and its role grew when the April 2026 interagency model-risk revision left generative and agentic AI to banks' broader risk programs: institutions now routinely present NIST-aligned Govern/Map/Measure/Manage programs to examiners as evidence of control over systems outside formal model-risk scope.

Where is the official NIST AI RMF 1.0 PDF, and what does the framework contain?

The official NIST AI Risk Management Framework 1.0 is document NIST AI 100-1, published January 26, 2023, and the PDF is hosted by NIST at nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf (DOI 10.6028/NIST.AI.100-1). It has two parts. Part 1 frames AI risk — how it differs from traditional software risk, who the intended audience is, and the seven characteristics of trustworthy AI: valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed. Part 2 is the Core: four functions — Govern, Map, Measure and Manage — each broken into categories and subcategories of outcomes that an organization tailors into a 'profile' for its own context. The framework is voluntary. Its companions are the online AI RMF Playbook, a Roadmap, crosswalks to other standards, and the Generative AI Profile (NIST AI 600-1, July 26, 2024), also a NIST-hosted PDF. NIST states that AI RMF 1.0 is being revised under the July 2025 White House AI Action Plan; no revised draft has been published.

The Core is what banks actually implement. Govern is cross-cutting: policies, accountability, workforce, culture and the processes that make the other three functions run. Map establishes context — the intended purpose, the deployment setting, the affected people, the risks and benefits — before a system is built or bought. Measure covers the metrics, tests and tracking that show whether the trustworthiness characteristics are met, including how the system is evaluated over time. Manage is the prioritisation and response: which risks are treated, transferred, avoided or accepted, and how incidents and decommissioning are handled. Each function's categories and subcategories are written as outcomes rather than controls, which is why the Playbook exists — it turns each outcome into suggested actions and the documentation an examiner would expect to see.

Profiles are the mechanism for sector-specific use. A profile is a selection and tailoring of Core outcomes for a use case, sector or technology; the Generative AI Profile is the first NIST-authored one and lists twelve risks — among them confabulation, data privacy, information security, harmful bias, intellectual property and value-chain integration — with actions mapped back to the Core subcategories. For a bank, an internal profile is typically the artefact that ties the framework to model risk management, third-party risk and information security programmes: the same Govern outcomes appear in each, and the Measure outcomes are where model validation evidence is filed.

Two practical notes on the documents themselves. The framework is voluntary and non-sector-specific; nothing in it creates an obligation, and US banking agencies have not incorporated it into guidance, though examiners increasingly expect something equivalent for generative and agentic AI that the April 2026 interagency model risk guidance leaves outside formal validation scope. And it is about to change: NIST says AI RMF 1.0 is being revised as part of the White House AI Action Plan of July 2025. Programmes built on the 1.0 Core will need a re-mapping when the revision lands; the Playbook will be updated after it.

WHAT THIS MEANS IN PRACTICE

• Cite the document, not a summary: NIST AI 100-1 (DOI 10.6028/NIST.AI.100-1) for the framework and NIST AI 600-1 for the generative AI profile — both PDFs are on nvlpubs.nist.gov and linked above.
• Build a written profile: select the Core subcategories that apply, record the tailoring, and map each to the bank's existing controls — that document is what examiners can read.
• Use the Playbook's suggested actions as the evidence checklist for each subcategory; it is the closest thing to an implementation guide NIST provides.
• Track the revision: when the revised framework is published, re-map the profile before the next examination cycle rather than after.

Which banks' AI programmes does NIST AI RMF 1.0 reach?

4 of the 100 largest US banks profiled on this site cite NIST AI RMF 1.0 among the documents their AI work answers to.