VA evidence errors on veterans' claims rose about fivefold as automation spread, report finds

The Department of Veterans Affairs (VA) is leaning harder on automation to decide disability claims, and its errors in collecting the evidence behind those claims have jumped, according to an investigation published Thursday ¹. The VA counted more than 66,000 evidence-collection errors in fiscal year 2021. In fiscal year 2026, the count was over 329,000 ¹.

The report came from The Beam, an investigative newsroom at Arizona State University's Howard Center for Investigative Journalism, and Military Times republished it ¹. It also found that higher-level VA reviewers are sending back more decisions. They returned about 23% in fiscal year 2021, just before the automated tool launched, and 35% in fiscal year 2026 ¹.

For a veteran, a returned or wrong decision usually means more waiting. Disability compensation is the monthly, tax-free payment the VA makes for injuries and illnesses connected to military service. When a claim is denied by mistake, it can go into an appeals process that takes years.

What the tool does

The software is called Automated Decision Support, or ADS. The VA introduced it in December 2021, starting with high blood pressure (hypertension) claims in Boise, Idaho ¹. It gathers medical records, decides whether a claim is ready for review, and writes summaries for the human employees who rate claims ¹.

Its use has grown fast. The VA made nearly 300,000 decisions with ADS in 2023, and by April 2026 the tool was acting on claims more than 50,000 times a month ¹. The department plans to use it on 90% of claims by the end of 2027 ¹. The proposed budget for fiscal year 2027 includes $130 million for automation and artificial intelligence (AI) at the Veterans Benefits Administration, the part of the VA that handles benefits ².

VA officials say people, not software, make the final call. At a House hearing in April, Margarita Devlin, the VA's principal deputy under secretary for benefits, said the AI "does not make any decisions and will not deny a claim." ²

The VA's defense

The VA disputes the investigation's central finding. Press secretary Quinn Slaven said the rise in errors came from appeals, which quadrupled from 2020 to 2026, not from the tool ¹.

There is no evidence to suggest the Automated Decision Support tool is causing errors

The department also points to speed. Devlin testified that average processing time fell 42% since January 2025, from 141 days to 81 days ². The VA reported that its backlog was under 100,000 claims, the first time that has happened since 2020 ². Devlin said the VA's accuracy rate was above 94% at the end of March ².

That number counts each separate medical issue in a claim. Rep. Tim Kennedy, a New York Democrat, noted that accuracy measured by whole claims was 83.31% ². Sandra Flint, a VA deputy under secretary, defended the issue-based measure ².

Warnings piled up

The concerns are not new. In September 2023, the VA's independent watchdog, the Office of Inspector General (OIG), reviewed the hypertension pilot. It found inaccurate and inconsistent determinations in 27% of the claims it checked, 16 of 60, and said the inaccuracies were largely caused by confusing guidance ⁴.

In April 2026, the inspector general looked at a separate automated system that grants benefits to surviving family members ⁵. It found that "At least 8,000 decisions or letters omitted favorable findings," and that at least 2% of decisions had legal errors, leading to an estimated $2.7 million in improper payments ⁵. The watchdog also said quality review for automated claims "was less rigorous than for traditional claims." ⁵

One veteran in the investigation, former Marine Randy Leger Diaz, was first denied most of his claims ¹. After a higher-level review, the VA admitted it had made evidence-collection errors, and he later received compensation for some of them ¹.

Democrats press on speed over accuracy

Rep. Mark Takano of California, the top Democrat on the House Veterans' Affairs Committee, has warned for months that the VA is putting speed ahead of getting claims right ³. At the April hearing, he said lawmakers were hearing that "claim processing errors are rising and denial rates are increasing." ³

Premature or erroneous denials lead to more appeals, where timelines are measured in years, and not months.

Takano also tied the problem to the department's workforce, pointing to "the toll that mass firings and a return to mandatory overtime have taken." ³ A Democratic committee aide told The Beam the rise in errors reflects a quality deficit that can be traced to the tool ¹.

Republicans on the committee have largely backed the technology. Rep. Morgan Luttrell, a Texas Republican, gave the receipt at the April hearing ².

we are eagerly behind the injection of AI models into the system so our veterans get what they need in a timely manner.

That matters because the balance of power is shifting. Democrats are preparing to use their new congressional majority to investigate the administration, and they could use subpoena power to do it ⁶. The VA's claims error numbers give them a concrete place to start: whether a goal of 90% automation by 2027 ¹ is being met by shortchanging the veterans it is supposed to serve.

The Record

1. Sep 25, 2023The VA inspector general reports that the automated hypertension claims pilot produced inconsistent decisions in 27% of claims it reviewed ⁴.
2. Apr 15, 2026At a House hearing, Rep. Mark Takano warns that claim errors and denials are rising as the VA focuses on speed ³.
3. Apr 30, 2026The inspector general finds that at least 8,000 automated survivor-benefit decisions or letters left out favorable findings, with an estimated $2.7 million in improper payments ⁵.
4. Nov 5, 2026An investigation finds the VA's evidence-collection errors rose from more than 66,000 in fiscal 2021 to over 329,000 in fiscal 2026 ¹.

Free to reprint Republish this story on your site or in your newsletter

Any site, blog or newsletter may run this story free of charge. Keep the credit line and its link to the original, do not sell it, and edit only for length or house style (the full terms). Photos carry their own licenses, noted under each photo on this page, so use the text alone unless a credit says otherwise.

<h2>VA evidence errors on veterans&#39; claims rose about fivefold as automation spread, report finds</h2> <p><em>By The GOP Files National Desk. This article first appeared on <a href="https://thegopfiles.com/story/va-evidence-errors-on-veterans-claims-rose-about-fivefold-as-automation-spread-report">The GOP Files</a> on Thursday, November 5, 2026.</em></p> <p>The Department of Veterans Affairs (VA) is leaning harder on automation to decide disability claims, and its errors in collecting the evidence behind those claims have jumped, according to an investigation published Thursday . The VA counted more than 66,000 evidence-collection errors in fiscal year 2021. In fiscal year 2026, the count was over 329,000 .</p> <p>The report came from The Beam, an investigative newsroom at Arizona State University&#39;s Howard Center for Investigative Journalism, and Military Times republished it . It also found that higher-level VA reviewers are sending back more decisions. They returned about 23% in fiscal year 2021, just before the automated tool launched, and 35% in fiscal year 2026 .</p> <p>For a veteran, a returned or wrong decision usually means more waiting. Disability compensation is the monthly, tax-free payment the VA makes for injuries and illnesses connected to military service. When a claim is denied by mistake, it can go into an appeals process that takes years.</p> <h2>What the tool does</h2> <p>The software is called Automated Decision Support, or ADS. The VA introduced it in December 2021, starting with high blood pressure (hypertension) claims in Boise, Idaho . It gathers medical records, decides whether a claim is ready for review, and writes summaries for the human employees who rate claims .</p> <p>Its use has grown fast. The VA made nearly 300,000 decisions with ADS in 2023, and by April 2026 the tool was acting on claims more than 50,000 times a month . The department plans to use it on 90% of claims by the end of 2027 . The proposed budget for fiscal year 2027 includes $130 million for automation and artificial intelligence (AI) at the Veterans Benefits Administration, the part of the VA that handles benefits .</p> <p>VA officials say people, not software, make the final call. At a House hearing in April, Margarita Devlin, the VA&#39;s principal deputy under secretary for benefits, said the AI &quot;does not make any decisions and will not deny a claim.&quot;</p> <h2>The VA&#39;s defense</h2> <p>The VA disputes the investigation&#39;s central finding. Press secretary Quinn Slaven said the rise in errors came from appeals, which quadrupled from 2020 to 2026, not from the tool .</p> <blockquote><p>There is no evidence to suggest the Automated Decision Support tool is causing errors</p><p>&mdash; Quinn Slaven, VA press secretary</p></blockquote> <p>The department also points to speed. Devlin testified that average processing time fell 42% since January 2025, from 141 days to 81 days . The VA reported that its backlog was under 100,000 claims, the first time that has happened since 2020 . Devlin said the VA&#39;s accuracy rate was above 94% at the end of March .</p> <p>That number counts each separate medical issue in a claim. Rep. Tim Kennedy, a New York Democrat, noted that accuracy measured by whole claims was 83.31% . Sandra Flint, a VA deputy under secretary, defended the issue-based measure .</p> <h2>Warnings piled up</h2> <p>The concerns are not new. In September 2023, the VA&#39;s independent watchdog, the Office of Inspector General (OIG), reviewed the hypertension pilot. It found inaccurate and inconsistent determinations in 27% of the claims it checked, 16 of 60, and said the inaccuracies were largely caused by confusing guidance .</p> <p>In April 2026, the inspector general looked at a separate automated system that grants benefits to surviving family members . It found that &quot;At least 8,000 decisions or letters omitted favorable findings,&quot; and that at least 2% of decisions had legal errors, leading to an estimated $2.7 million in improper payments . The watchdog also said quality review for automated claims &quot;was less rigorous than for traditional claims.&quot;</p> <p>One veteran in the investigation, former Marine Randy Leger Diaz, was first denied most of his claims . After a higher-level review, the VA admitted it had made evidence-collection errors, and he later received compensation for some of them .</p> <h2>Democrats press on speed over accuracy</h2> <p>Rep. Mark Takano of California, the top Democrat on the House Veterans&#39; Affairs Committee, has warned for months that the VA is putting speed ahead of getting claims right . At the April hearing, he said lawmakers were hearing that &quot;claim processing errors are rising and denial rates are increasing.&quot;</p> <blockquote><p>Premature or erroneous denials lead to more appeals, where timelines are measured in years, and not months.</p><p>&mdash; Mark Takano, Ranking member, House Veterans&#39; Affairs Committee, D-Calif.</p></blockquote> <p>Takano also tied the problem to the department&#39;s workforce, pointing to &quot;the toll that mass firings and a return to mandatory overtime have taken.&quot; A Democratic committee aide told The Beam the rise in errors reflects a quality deficit that can be traced to the tool .</p> <p>Republicans on the committee have largely backed the technology. Rep. Morgan Luttrell, a Texas Republican, gave the receipt at the April hearing .</p> <blockquote><p>we are eagerly behind the injection of AI models into the system so our veterans get what they need in a timely manner.</p><p>&mdash; Morgan Luttrell, U.S. representative, R-Texas</p></blockquote> <p>That matters because the balance of power is shifting. Democrats are preparing to use their new congressional majority to investigate the administration, and they could use subpoena power to do it . The VA&#39;s claims error numbers give them a concrete place to start: whether a goal of 90% automation by 2027 is being met by shortchanging the veterans it is supposed to serve.</p> <p><em>Read the original, with every source linked, at <a href="https://thegopfiles.com/story/va-evidence-errors-on-veterans-claims-rose-about-fivefold-as-automation-spread-report">thegopfiles.com</a>. Republished under The GOP Files' <a href="https://thegopfiles.com/republish">free republishing terms</a>.</em></p> <img src="https://thegopfiles.com/api/rp/va-evidence-errors-on-veterans-claims-rose-about-fivefold-as-automation-spread-report.gif" alt="" width="1" height="1" style="border:0;width:1px;height:1px">

Copy the HTML Paste it into your editor's HTML view. The one-pixel image at the end tells our newsroom the story ran; it stores no reader data.

Sources

1. Military Times (The Beam, Howard Center for Investigative Journalism), “As the VA leans on automation, errors in veterans' claims soar” (2026-11-05). militarytimes.com/news/2026/11/05/as-the-va-leans-on-automat
2. Nextgov/FCW, “AI is helping VA speed up claims processing, but Dems worry about errors” (2026-04-16). nextgov.com/artificial-intelligence/2026/04/ai-helping-va-sp
3. House Veterans' Affairs Committee Democrats, “Ranking Member Takano's Remarks at Hearing on VA Disability Claims Backlog” (2026-04-15). democrats-veterans.house.gov/news/press-releases/ranking-mem
4. VA Office of Inspector General, “Improvements Needed for VBA's Claims Automation Project (Report 22-02936-175)” (2023-09-25). vaoig.gov/node/19593
5. VA Office of Inspector General, “Review of Automated Decisions for Veterans' Service-Connected Death Claims (Report 25-00153-47)” (2026-04-30). vaoig.gov/node/22762
6. ABC News (via ABC7 San Francisco), “President Donald Trump dismisses prospect of replacing Defense Secretary Pete Hegseth amid Iran war fallout, looming Democratic investigations” (2026-11-05). abc7news.com/story/president-donald-trump-dismisses-prospect

Found an error? Tell the newsroom. Corrections are noted at the top of the story.