Governance & ComplianceUpdated 8 min read

NIST AI RMF vs. EU AI Act in Healthcare & Banking: A Practical Compliance Checklist

The short answer

NIST AI RMF 1.0 is voluntary guidance for managing AI risk. The EU AI Act is binding law, with duties that depend on scope, intended use, classification, role, and applicable provisions. A shared control library can organize evidence, but adopting NIST or completing this checklist does not establish legal compliance. Assess each system separately.

Two frameworks, one governance program

Compliance and engineering leaders in regulated industries keep asking the same question: do we need a NIST program and a separate EU AI Act program? Our suggested approach is to reuse relevant controls where appropriate while maintaining a separate, requirement-by-requirement legal assessment. Reuse is a practical option, not evidence that all obligations are met.

This matters most in healthcare and financial services, where AI already sits on top of existing privacy, financial, medical-device, and supervisory obligations. Neither NIST adoption nor an EU AI Act review replaces a separate assessment of those requirements. Confirm the current rules and their applicability with qualified advisers.

Side by side: NIST AI RMF and the EU AI Act

DimensionNIST AI RMFEU AI Act
Legal statusVoluntary U.S. frameworkBinding EU regulation
StructureFour functions: Govern, Map, Measure, ManageProhibitions, high-risk requirements, transparency duties, and general-purpose AI rules
FocusHow to manage AI risk across the lifecycleWhat obligations apply to each type of AI system
Who it applies toOrganizations choosing to use itActors within Article 2 scope; obligations differ by role and system
EvidenceSuggested risk outcomes and documented evaluationsRole-specific records; high-risk provider and deployer duties differ
Best used asA way to organize risk-management outcomesA legal requirements assessment, not replaced by NIST adoption

Using NIST functions to organize a legal gap review

This is an editorial organizing approach, not an official crosswalk or a statement of equivalence. NIST functions are interconnected; one legal obligation may involve several functions.

  • Govern — organize accountability, policy, and training evidence; separately assess applicable organizational and AI literacy duties.
  • Map — document context and intended use; separately assess territorial scope, legal role, and classification under Articles 2 and 6.
  • Measure — organize evaluation evidence; separately assess applicable high-risk data-governance, accuracy, robustness, and cybersecurity requirements.
  • Manage — prioritize risk treatments and review outcomes; separately assess applicable oversight, logging, monitoring, and reporting duties.

Where healthcare and banking feel it most

Annex III, points 5(b) and 5(c) includes creditworthiness evaluation and credit scoring of natural persons, except financial-fraud detection, and risk assessment and pricing concerning natural persons in life and health insurance. Article 6(3) sets conditions for certain Annex III exceptions; profiling of natural persons remains high-risk.

For medical-device-related AI, Article 6(1) requires both an Annex I product or safety-component connection and required third-party conformity assessment under the relevant product law. Clinical support is not automatically high-risk just because it involves healthcare. Internal search or drafting is not automatically exempt either: intended use, data, scope, and applicable transparency or other duties still need review.

A 10-point AI compliance checklist

Use these suggested governance checks to start a gap review, not as an exhaustive legal checklist. For high-risk systems, distinguish provider obligations in Articles 8–25 from deployer obligations in Article 26; other provisions may also apply. Confirm current application dates and any amendments separately.

  • Keep a live inventory of every AI system, including vendor tools and embedded features.
  • Assess intended use, Article 2 scope, provider/deployer role, and Article 6 classification; record the reasoning and exceptions.
  • Name an accountable business owner and a technical owner for every production system.
  • Document intended use, data sources, and known limitations before go-live.
  • Identify applicable privacy and sector requirements; test authorized data access, including retrieval permissions, against documented policies.
  • Test for accuracy, bias, and robustness against a documented baseline before release.
  • Define appropriate human oversight, competence, intervention rights, and records for the intended use and applicable duties.
  • Set logging and retention rules proportionate to risk and legal duties; protect sensitive data rather than recording every input indiscriminately.
  • Set risk-based evaluation and monitoring schedules, incident escalation, and any applicable reporting duties.
  • Review vendor AI contracts for data use, audit rights, and change notification.

Completing these checks does not establish conformity, certification, or legal compliance. Use the findings to assign evidence gaps and accountable owners. Our AI Governance Maturity Model offers a suggested internal evidence rubric, not an official NIST score or legal assessment. This article is general information, not legal advice.

Frequently asked questions

What is the difference between the NIST AI RMF and the EU AI Act?

NIST AI RMF 1.0 is a voluntary risk-management framework organized around Govern, Map, Measure, and Manage. The EU AI Act is binding EU legislation with obligations that depend on territorial scope, intended use, system classification, the organization's role, and applicable provisions. Using NIST does not establish EU AI Act compliance.

Is the NIST AI RMF mandatory for healthcare or banking?

NIST AI RMF 1.0 is voluntary; it is not itself a healthcare or banking law. An organization may adopt it through internal policy or contractual commitments. Applicable privacy, financial, medical-device, and supervisory requirements must be assessed separately; an RMF assessment does not demonstrate compliance with those requirements.

Does the EU AI Act apply to U.S. healthcare and financial companies?

It can. Article 2 covers, among others, providers placing AI systems on the EU market or putting them into service in the EU, deployers established or located in the EU, and third-country providers or deployers where the system's output is used in the EU. Exceptions and the organization's role matter. EU patients or customers alone are not a complete scope test; obtain a system-specific legal assessment.

Which AI uses in healthcare and banking are likely high-risk under the EU AI Act?

Annex III includes AI used to evaluate natural persons' creditworthiness or establish credit scores, except systems used to detect financial fraud, and AI for risk assessment and pricing concerning natural persons in life and health insurance. Under Article 6(1), product-related AI must meet both the Annex I product or safety-component condition and the third-party conformity-assessment condition. Not every healthcare AI system is automatically high-risk. Assess Article 6, its conditions and exceptions, and the specific intended use.

Can one governance program satisfy both frameworks?

A shared control library can help organize overlapping risk-management work, but it does not automatically satisfy the EU AI Act. Map each applicable legal requirement to the responsible role, evidence, reviewer, and gap. High-risk provider duties and deployer duties differ; additional conformity assessment, registration, documentation, oversight, monitoring, or reporting may apply. This checklist is not a legal compliance determination.

Official sources and scope

Sources checked November 4, 2026. The NIST mapping and checklist above are editorial guidance, not an official legal crosswalk. Requirements and application dates may change; verify current law and system-specific obligations before deployment.

See how your AI program measures up

Our AI Readiness Assessment reviews your governance, data, and operating controls against frameworks like NIST AI RMF and gives you a sequenced plan to close the gaps.