GAO Warns VA’s Push to Use AI for Disability Claims Faces Longstanding IT and Oversight Gaps

A new Government Accountability Office testimony says the Department of Veterans Affairs is trying to modernize the technology behind disability claims and explore artificial intelligence for a benefits program of enormous scale, even as longstanding management and IT weaknesses remain unresolved and a recent automation breakdown underscored the risks.

The testimony, titled “VA Disability Benefits: Opportunities and Challenges to Modernizing Technology and Adopting AI,” was publicly released Friday ahead of planned testimony Monday before the House Committee on Veterans’ Affairs’ Subcommittee on Technology Modernization. The GAO said the VA paid more than $195 billion in disability compensation to more than 6.9 million veterans and their families in fiscal 2025, making the stakes of any technology changes unusually high.

Dr. Sterling Thomas, chief scientist in the GAO’s Science, Technology Assessment, and Analytics team, prepared the statement, identified as GAO-26-109137. The PDF says it is “For Release on Delivery Expected at 3:00 p.m. ET Monday, July 13, 2026.” GAO said, “VA is working to modernize its IT systems to increase efficiency and accuracy of claims processing for disability compensation.”

But the watchdog’s central warning is that the department is not starting from a clean slate. The testimony draws on GAO work from September 2015 through June 2026 and says the VA still has long-standing problems in running the disability compensation program and carrying out technology upgrades.

Among the issues GAO highlighted are gaps in the VA’s oversight of the quality of medical exams performed by contracted providers, shortcomings in how it manages training for claims processors, and weak planning in earlier work to modernize the Veterans Benefits Management System, the core claims-processing platform. That modernization effort began in 2009, but GAO said it was not driven by robust planning and lacked goals for system response times and user satisfaction, making progress difficult to measure.

Since 2021, GAO said, it has made 43 recommendations to improve the disability compensation program. Twenty-eight have been implemented, while 15 remain not fully implemented.

The testimony says AI could help government operations and notes that the VA is exploring ways to use AI to further automate disability claims processing. That push is also part of a broader policy mandate: The Honoring our PACT Act of 2022 required the VA to develop a plan to modernize its benefits IT systems to improve claims processing, and the department’s 2024 addendum to its five-year modernization plan included AI and automation efforts aimed at gathering, structuring and preprocessing claims evidence.

Still, GAO warned that newer AI tools can create fresh oversight problems. Referring to generative AI, the testimony says it “can increase risk and hinder accountability, in part because even its designers may not fully understand how it works.” GAO also said such systems can require significant computational and technical resources and can make it harder to detect errors or misuse because of limited transparency.

The testimony points to a recent VA oversight case as a practical example of automation risk in benefits administration. GAO cited an April 30, 2026, review by the VA Office of Inspector General, the department’s internal watchdog, of automated decisions for service-connected death claims.

That review “found legal and procedural deficiencies in automated rating decisions and notification letters,” the inspector general said. According to GAO’s summary, at least 8,000 decisions or letters omitted favorable findings or had other issues, and at least 2% contained legal errors, resulting in an estimated $2.7 million in improper payments.

GAO did not present that case as evidence that the VA is broadly using generative AI to decide disability claims. Instead, the testimony uses it to show that automation in veterans’ benefits can produce consequential errors if oversight, transparency and controls fall short.

As the VA moves ahead with technology modernization required under the PACT Act, GAO said the department and other agencies could use the watchdog’s AI Accountability Framework when selecting and implementing AI systems.