Standardizing AI Model Transparency in U.S. Federal Agencies: A Framework for Implementing AI Model Cards and Data Provenance Auditing

Résumé fourni par la source

As artificial intelligence becomes deeply embedded in U.S. federal agency operations spanning healthcare delivery, law enforcement, benefits administration, national security, and regulatory rulemaking the imperative for systematic, standardized transparency has never been more pressing.This article examines the landscape of AI model transparency in U.S. federal agencies, synthesizing recent policy developments including OMB Memoranda M-25-21 and M-25-22 (April 2025), Executive Order 14179, the December 2025 OMB "Unbiased AI" guidance, the NIST AI Risk Management Framework (AI RMF), and the evolving doctrine of data provenance auditing.We analyze the 2024 Federal AI Use Case Inventory which disclosed more than 2,133 use cases including 227 rights-and safety-impacting deployments and evaluate the current patchwork of transparency requirements against emerging best practices for AI model cards.We then propose a comprehensive, five-tier standardization framework for implementing model cards and data provenance auditing across federal agencies, including governance structures, technical standards, enforcement mechanisms, and workforce capacity requirements.Our findings reveal significant inconsistencies in current reporting, dangerous gaps in thirdparty auditing capacity, and a critical need for interoperable provenance infrastructure.We conclude with policy recommendations for Congress, OMB, NIST, and Chief AI Officers.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.