Government Agencies Acquire AI As They Go and Without Any Regulatory Framework

Why it Matters

AI procurement federal agencies rely on has more than doubled in just one year, and yet not one of the four major agencies reviewed by the Government Accountability Office had a formal policy requiring officials to document what worked, what failed, and what cost more than expected. The result: billions of dollars in contracts being negotiated without institutional memory, standardized terms, or the technical expertise needed to evaluate what contractors are actually selling.

The GAO report, published April 13, 2026, finds that agencies are largely improvising, repeating mistakes that a shared knowledge base could prevent, and signing contracts built on regulatory frameworks designed before artificial intelligence existed.

The Scale of the Problem

Federal AI use more than doubled between 2023 and 2024, with agencies across the government acquiring AI capabilities through fiscal year 2025. The Department of Defense, the Department of Homeland Security, the General Services Administration, and the Department of Veterans Affairs were the four agencies GAO examined directly.

Their uses of AI span a wide range of government functions. DHS deploys AI for facial recognition at airports, the VA uses it to analyze veterans' benefit claims, and the DOD has integrated AI into defense and logistics systems. These are operational systems touching millions of Americans.

And yet the procurement infrastructure supporting these systems remains fragmented, underdeveloped, and largely unaccountable.

Six Ways AI Vendor Selection Challenges Are Going Unaddressed

GAO identified six recurring procurement challenges that agencies are encountering and failing to systematically address:

1. Lack of technical expertise. Agencies reported difficulty accessing AI technical experts capable of evaluating contractor proposals. Without that expertise in-house, the government is often in a weaker negotiating position than the vendors it is hiring.

2. Cost opacity. Officials said it was hard to understand AI-related costs. Unlike traditional IT procurement, AI systems involve complex pricing structures that do not map neatly onto existing cost-estimation frameworks.

3. No standardized contract language. There are no widely adopted standard contract terms for AI acquisitions. Each agency is largely writing its own terms, creating inconsistency and legal vulnerability.

4. Software license complications. AI systems often come bundled with software licenses that carry their own restrictions and costs, another layer of complexity that procurement officers are navigating without formal guidance.

5. No institutional memory. When an acquisition officer completes an AI contract, that knowledge largely disappears. There is no systematic mechanism for capturing what was learned and passing it on.

6. Outdated contract frameworks. The Federal Acquisition Regulations, the government's primary contracting rulebook, were designed for a pre-AI era. Agencies are trying to fit AI acquisitions into a framework that was not built for them.

Lessons Learned?

The GAO's most pointed finding is the absence of any formal lessons-learned process across all four agencies studied. Officials are "figuring out AI acquisition on their own," according to reporting by FedScoop, without the benefit of what other agencies have already discovered through trial and error.

The Office of Management and Budget had already moved to address this gap. In April 2025, OMB issued guidance implementing Executive Order 14179, directing agencies to improve AI governance and share acquisition knowledge through a web-based repository managed by GSA. The problem: agencies are not yet feeding into it. The infrastructure exists. The compliance does not.

GAO's core recommendation is direct: DOD, DHS, GSA, and VA should update their internal policies to require officials to systematically collect and share lessons learned from AI acquisitions. All four agencies are on record as having received the recommendation. Whether they act on it is another matter.

Federal Agency AI Implementation and the Oversight Gap

The report arrives at a moment when AI procurement federal agencies conduct is drawing scrutiny from multiple directions. The Trump administration's April 2025 OMB memoranda pushed agencies to accelerate AI adoption and establish governance frameworks. GAO's findings suggest the governance side has not kept pace with the adoption side.

When agencies lack AI technical expertise procurement processes to evaluate contractor proposals, they are relying on vendors to define the terms of their own oversight. When there is no standardized contract language, each negotiation starts from scratch and costs time and money, potentially exposing the government to unfavorable terms.

The Federal Acquisition Regulations, cited in the GAO report as an inadequate framework for modern AI contracting, have not been updated to reflect the realities of artificial intelligence acquisitions government-wide. That regulatory lag compounds every other challenge agencies face.

The Bottom Line

The recommendations in the report are procedural rather than sweeping. GAO is not calling for a moratorium on AI procurement or a wholesale restructuring of federal contracting law. It is asking four agencies to write down what they learn and share it.

Specifically, GAO recommends that DOD, DHS, GSA, and the VA each update their acquisition policies to require the systematic collection and dissemination of lessons learned from AI procurements. The GAO also recommends that GSA's existing OMB-directed repository be actually used for that purpose.

Access the Legis1 platform for comprehensive political news, data, and insights.