Federal Agencies Are Buying AI Fast—and Making Expensive Mistakes
A new report from the Government Accountability Office arrives at a moment when federal AI spending is accelerating faster than the institutional frameworks meant to govern it. Released April 13, 2026, GAO-26-107859 examines how four major agencies—the Department of Defense, the Department of Homeland Security, the General Services Administration, and the Department of Veterans Affairs—have been acquiring AI capabilities, and finds a consistent pattern: agencies are learning hard lessons in isolation, then failing to share what they’ve learned.
The headline finding is structural rather than technical. Federal agencies more than doubled their use of AI between 2023 and 2024, according to the Federal Chief Information Officer, and Congress recently appropriated approximately $1.7 billion for AI efforts across government. But the procurement machinery supporting that expansion is visibly straining. Agency officials across all five organizations interviewed for the report told GAO they essentially had to figure out AI acquisition on their own—without institutional memory, without standardized contract language, and without a systematic way to communicate what worked and what didn’t to their counterparts elsewhere in government.
The Office of Management and Budget issued Memorandum M-25-22 in April 2025 specifically to address this gap, directing agencies to share knowledge about AI acquisitions through a web-based repository managed by GSA. The problem, GAO found, is that none of the four agencies in scope had policies requiring officials to actually collect lessons learned in the first place. You cannot share what you have not documented.
The report covers thirteen AI acquisitions in depth—ranging from facial recognition systems at TSA to generative AI chatbots at the Air Force to seizure-detection deep learning tools at VA—and uses them to illustrate six recurring challenge areas that cut across strategic and programmatic lines. Those challenges include access to subject matter experts, data and intellectual property rights protections, acquisition timelines misaligned with AI development cycles, requirements definition, testing and continuous evaluation, and pricing opacity.
All four agencies concurred with GAO’s recommendations, which call on each to update departmental policies to require systematic collection and submission of AI acquisition lessons learned to the GSA-managed repository. DHS set a target completion date of July 31, 2026; VA set August 1, 2026.
The full report is available at gao.gov under report number GAO-26-107859.