Opinion articles provide independent perspectives on key community issues, separate from our newsroom reporting.

Opinion

The VA is using AI for claims now. Veterans deserve human trust | Opinion

The Veterans Administration is sidestepping scrutiny by calling much of it “automation.” That word sounds harmless. It is not.
The Veterans Administration is sidestepping scrutiny by calling much of it “automation.” That word sounds harmless. It is not. Getty Images

The last thing a veteran who gets a mistaken benefits denial can expect is an apology. What usually comes instead is decision delays: months, sometimes years, and time lost that cannot be made up even with a win after waiting decades for an appeal victory. Every year, tens of thousands of veterans are forced into that grind because the system meant to serve them still fails to work as promised.

Now, to make things worse, the U.S. Department of Veterans Affairs is using artificial intelligence-driven and adjacent technologies to process claims, influence adjudication outcomes and handle financial transactions affecting millions of veterans, while sidestepping scrutiny by calling much of it “automation.” That word sounds harmless. It is not.

In a 2025 VA Privacy Impact Assessment, the agency describes its automation platform using robotic process automation, optical character recognition, natural language processing and AI to process veterans’ benefits information and support adjudication. That’s AI in everything but the press release. By avoiding the label, VA can call the result an “automated error,” which doesn’t carry the same weight as “AI error.” It’s harder to investigate, harder to report on and harder for veterans to challenge when they don’t know what they’re fighting against.

These automated errors have already surfaced in VA Office of Inspector General reports. In 2023, the OIG reported that the VA’s claims automation project produced inaccurate or inconsistent outcomes in 27% of reviewed claims — more than 1 in 4. In April 2025, OIG found that the VA used two tools with significant shortcomings on decisions under the PACT Act for veterans exposed to toxic substances, estimating that approximately 31,400 claims were decided improperly, and warning of at least $20.4 million in additional improper payments if the error rate continued.

Then, earlier this year, the OIG warned that the VA had rolled out generative AI without adequate safeguards for patient safety and veteran data, cautioning these systems can produce inaccurate outputs capable of affecting clinical decisions. The risk was serious enough that OIG issued its first-ever Preliminary Result Advisory Memorandum, an extraordinary move signaling the danger could not wait for the usual slow drip of oversight. Inspector General Cheryl Mason warned that the VA moved too quickly to deploy AI without first building the oversight needed to catch mistakes.

Veterans and the public hear “automation” and picture a back-office process that folds a letter into a pre-stamped envelope. What the VA is often describing instead is a stack of machine-driven systems — such as robotic process automation, optical character recognition, natural language processing and machine learning — that shape high-stakes decisions without meaningful human review or a reliable system to catch mistakes.

The Veterans Health Administration is willing to say “artificial intelligence,” while the Veterans Benefits Administration hides behind “automation” — but no matter the name, machine-driven errors in benefits processing can still distort payments, multiply across thousands of claims and harm the livelihoods of those that risked life and limb for their nation.

Veterans served this country with the understanding that if they were harmed in that service, the government would keep its promise. When the VA hands life-changing health care decisions to systems that miscalculate payments, distort ratings or deny earned relief, it breaks that promise.

When AI or AI-adjacent systems contribute to an error, reports and public documentation need to say so plainly. Any machine-driven system that touches benefits adjudication should require mandatory human review before a negative outcome becomes final. Veterans harmed by those systems deserve proactive notice and remediation, not just relief if they survive a yearslong appeal. And veterans should know when AI played a role in the process. If a machine helped shape the decision, the veteran has a right to know.

Sitting across from veterans who spent years fighting decisions that were wrong from the beginning, the one thing most had in common was that they never learned why. The Veterans Administration owes veterans more than efficiency. It owes them honesty, accountability and a system worthy of their trust.

Benjamin Krause is a U.S. Air Force veteran, VA-accredited attorney, founder of DisabledVeterans.org and recognized veterans policy expert. He practices veterans benefits law at Krause Law, PLLC.

Get unlimited digital access
#ReadLocal

Try 1 month for $1

CLAIM OFFER