The Government Accountability Office is the independent, nonpartisan audit and evaluation agency of the US Congress. It conducts audits, investigations, and evaluations of federal programs and spending, and provides analyses, recommendations, and other assistance to Congress. In the AI domain, GAO audits whether federal agencies are implementing AI-related laws and executive orders and issues public recommendations when it finds agencies fall short.
Snapshot
Government-wide AI requirements cataloged
| Date | Requirements | Sources covered | Report |
|---|---|---|---|
| Jul 2025 | 94 current government-wide AI-related requirements | 5 laws, 6 EOs, 3 guidance documents | GAO-25-107933 (Sep 2025) |
AI-relevant work
Accountability Framework (2021)
In June 2021, GAO published Artificial Intelligence: An Accountability Framework for Federal Agencies and Other Entities (GAO-21-519SP). The framework describes four principles for responsible AI in government — governance, data, performance, and monitoring — with associated key practices and audit questions.
Federal AI requirements inventory (2023, 2025)
In December 2023 (GAO-24-105980), GAO reported that agency AI inventories included approximately 1,200 current and planned use cases. It made 35 recommendations to 19 agencies (including OMB) for fuller implementation of AI requirements, and found that OMB had not developed required guidance for AI acquisition, meaning agencies could not implement key inventory requirements.
In September 2025 (GAO-25-107933), GAO cataloged 94 current government-wide AI-related requirements across 5 laws, 6 executive orders, and 3 guidance documents as of July 2025. It found that only 4 of the 35 recommendations made in 2023 had been implemented — a roughly 11% implementation rate over approximately 18 months — and reported that OMB declined to respond to GAO's comment request. GAO's 2025 work characterizes the federal AI compliance landscape as having substantial requirements on paper alongside weak implementation.
These findings connect to broader patterns described elsewhere: traditional regulation with binding requirements but weak enforcement (Regulatory Typology: Self-Regulation, Co-Regulation, Traditional Government Regulation), the gap between what laws require and what agencies do (Regulating Under Uncertainty), and executive-agency tension distinct from state-federal tension (Techno-Federalism: How Regulatory Fragmentation Shapes the U.S.-China AI Race).
Environmental and human effects of generative AI (2025)
In GAO-25-107172, GAO reported on the environmental footprint (energy, water) and human effects (labor, content moderation) of generative AI. (Source: GAO-25-107172 — Generative AI's Environmental and Human Effects)
Leadership
Kevin Walsh serves as Director, Information Technology and Cybersecurity, and is named on the 2025 reports.
Relationships
- supports: Regulatory Typology: Self-Regulation, Co-Regulation, Traditional Government Regulation, Regulating Under Uncertainty
- supports: Executive Order 14179 — Removing Barriers to American Leadership in Artificial Intelligence, Executive Order 14319 — Preventing Woke AI in the Federal Government, America's AI Action Plan (as auditor of their implementation)
- related: National Institute of Standards and Technology (NIST), US AI Safety Institute (NIST AISI), OMB Memorandum M-24-10, OMB Memorandum M-24-18
- instance-of: oversight of GSA — General Services Administration (AI Deployer), DOD — Department of Defense (AI Deployer), DHS — Department of Homeland Security (AI Deployer)