The Department of Veterans Affairs is exploring artificial intelligence to speed and improve disability decisions, but federal auditors say unresolved management, training and technology weaknesses could limit the effort’s success.
By yourNEWS Media Newsroom
WASHINGTON — The Department of Veterans Affairs faces significant operational and technological obstacles as it considers using artificial intelligence to process disability claims, according to a new Government Accountability Office report that warns longstanding problems could complicate the agency’s modernization plans.
The VA is seeking to update technology used by the Veterans Benefits Administration in an effort to produce faster and more accurate disability determinations. However, Dr. Sterling Thomas, the GAO’s chief scientist, told lawmakers that unresolved weaknesses in program management and information technology remain a serious concern.
“VA is working to modernize its IT systems to deliver an improved veteran experience and increase efficiency and accuracy in claims processing,” Thomas said during testimony before the House Veterans’ Affairs Subcommittee on Technology Modernization.
“However, VA has faced longstanding challenges in managing programs and IT projects,” he added.
The disability compensation program is among the federal government’s largest benefit systems. During fiscal year 2025, the VA distributed more than $195 billion to more than 6.9 million veterans and family members.
That workload has continued to expand, partly because of the Honoring our PACT Act, which widened eligibility for veterans exposed to toxic substances during military service. The GAO said additional changes to eligibility rules in the future could generate further increases in the number of claims submitted to the department.
The challenge extends beyond the sheer volume of applications. Disability determinations often require processors to apply complicated and changing legal standards, regulations and agency policies to individual medical and service records.
“Making decisions about disability compensation is difficult given the complexity of VA’s statutory and policy requirements,” Thomas said.
The GAO report revisited several deficiencies previously identified by the watchdog, including inadequate oversight of medical examinations performed by contractors and weaknesses in the training provided to claims processors.
Since 2021, the GAO has issued 43 recommendations intended to improve the disability compensation program. The VA has fully carried out 28 of them, leaving 15 recommendations that have not yet been completely implemented.
Thomas said resolving those outstanding issues would put the department in a stronger position to update its technology and provide veterans with “more accurate and timely benefits decisions.”
The report also examined the VA’s previous record of technology modernization, pointing to the Veterans Benefits Management System as a central example.
The electronic claims platform was launched in 2009, but the GAO found that its development lacked sufficient planning and clear performance measures.
The system “was not driven by robust planning and did not include goals for system response times and user satisfaction,” the report said.
Without those benchmarks, auditors said, the VA had greater difficulty determining whether the platform was performing effectively or meeting the needs of employees and veterans.
Despite those concerns, the GAO acknowledged that properly managed artificial intelligence could improve government operations and reduce time spent on repetitive administrative work.
“AI holds substantial promise for improving the operations of government agencies,” Thomas said.
The VA is examining tools that could automate the intake of documents, assist in reviewing disability claims and identify potentially fraudulent changes to direct-deposit information connected to veterans’ benefits.
Such systems could reduce manual workloads and help employees process claims more efficiently. The GAO cautioned, however, that the department must address significant risks before relying on AI in decisions that directly affect veterans’ compensation.
Generative AI systems can function as “black boxes,” meaning the reasoning behind their conclusions may not be fully visible or understandable.
“Even the designers may not fully understand how they generate outputs,” the report warned.
That lack of transparency could make it more difficult for the VA to discover mistakes, assign responsibility for incorrect decisions or recognize improper use of an AI system.
The GAO also identified risks involving cybersecurity, privacy protections, computing capacity and access to complete, accurate and reliable data.
Auditors stressed that human review must remain part of the disability claims process even as the department introduces more automated technology. Keeping a “human in the loop,” the report said, is essential to reviewing AI-generated information and preventing automated errors from determining whether veterans receive benefits.
The agency will also need more employees with specialized technical knowledge to oversee, evaluate and maintain the systems.
The GAO said the VA and other federal departments already face difficulties recruiting and retaining workers with advanced AI expertise because government agencies must compete with private companies for a limited pool of qualified specialists.
The report concluded that artificial intelligence could help the VA manage its growing claims workload, but only if the department first strengthens oversight, completes outstanding reforms and avoids repeating weaknesses that affected earlier technology projects.
Original article
Comments