Department of Veterans Affairs Admits Using AI in Hundreds of High-Impact Systems
In the age of artificial intelligence, some veterans wonder just how much of the assistance they get from the Department of Veterans Affairs utilizes AI. The VA issued an audit that answers this question, reviewing a whopping 367 artificial intelligence systems currently operating across the VA.
Department of Veterans Affairs Using AI in Hundreds of High-Impact Systems
Of these 367 AI systems, 215 are classified as so-called “high-impact” operations, such as healthcare delivery, disability benefits processing, records management, and cybersecurity.
The VA report, called the VA AI Inventory, is, according to the federal government, “VA’s official record of AI use across the department,” and the VA claims, “It promotes transparency, supports responsible AI governance, and helps ensure that AI tools are aligned with Veteran needs and federal requirements.”
AI at the VA
Where has the VA deployed AI? Tools like the Ambient AI Scribe, designed to assist medical providers by listening to patient-clinician conversations, are just one example among many.
This system automatically generates clinical notes, an effort intended to reduce administrative burdens on staff and allow providers more face-to-face time with patients using the VA healthcare system. Other AI tools assist with medical imaging analysis, diagnostic workflows, and patient record summarization.
More than 50,000 personnel across the department now use commercial AI products like Microsoft Copilot and various transcription tools.
>> Never miss benefits news and updates that are important to you. Sign up today to receive the MyMilitaryBenefits newsletter free to your inbox.
Benefits and Processing Efficiency?
AI is increasingly integrated into the claims process. One high-impact tool, the TERA (Toxic Exposure Risk Activity) Memorandum Automation, assists claims processors by pre-populating answers on exposure forms using data already in a veteran’s file.
The AI-powered VA.gov chatbot is a high-visibility system that falls under the AI high-impact designation. This tool serves as a self-service interface, helping veterans find information, manage appointments, and access support services without manual intervention from a human representative.
How Safe is VA AI?
The Department of Veterans Affairs claims it requires (at press time) that every high-impact system get an evaluation before it is rolled out. This may include pre-deployment testing, assessments, independent reviews, and the creation of risk mitigation plans.
The VA wants us to believe these systems are subject to continuous monitoring, but the record of Department of Veterans Affairs failures with advanced computer systems is well-documented.
The Office of Inspector General has cautioned that the lack of standardized management for AI in clinical environments could present risks to patient safety. There are concerns about the potential for AI to produce inaccurate or hallucinated outputs, a common problem across AI platforms. which, if integrated into medical records without proper human verification, could lead to incorrect diagnostic or treatment decisions.
AI Industry Backing Away From Safety?
The VA transition to artificial intelligence use in its systems has not been without scrutiny, and with good reason, as some of the prime movers in the AI industry are “backing away from some safety protocols,” according to reporting by NPR and a third-party review of AI by an agency known as The Future of Life Institute.
That agency, according to NPR, says “…none of the companies ranked above a C+ in its latest AI Safety Index. Anthropic, OpenAI, Google DeepMind, and Meta were ‘moving the goalposts,’ the evaluators found, weakening previous commitments to safety measures.”
In light of this, are we really expected to believe the VA is up to the challenge? The GAO went on the record about AI use in a different government agency (the Department of Homeland Security), saying the agency’s audit of the Department of Homeland Security’s use of AI found “that DHS’s data aren’t always reliable.” Are we meant to believe the Department of Veterans Affairs AI use will be MORE reliable than the Department of Homeland Security?
Accountability and Oversight
The Government Accountability Office, which developed an AI Accountability Framework, says meaningful AI oversight requires effort; GAO representatives have argued that agencies must maintain records that independent third parties can follow to hold agencies accountable.
Veteran advocacy groups have concerns over AI versus the human element. The Veterans of Foreign Wars, for instance, says these technologies should serve as tools for human staff, not as replacements for human judgment.
There is a demand for accountability by these agencies, particularly in processes that affect a veteran’s disability rating or access to care. Advocates argue that the VA must commit to long-term AI monitoring. They want to make certain AI systems function correctly and prevent glitches or hallucinations from degrading performance or accuracy.
The future of AI at the Department of Veterans Affairs isn’t clear at press time.
>> Never miss benefits news and updates that are important to you. Sign up today to receive the MyMilitaryBenefits newsletter free to your inbox.
About the author
Editor-in-Chief Joe Wallace is a 13-year veteran of the United States Air Force and a former reporter/editor for Air Force Television News and the Pentagon Channel. His freelance work includes contract work for Motorola, VALoans.com, and Credit Karma. He is co-founder of Dim Art House in Springfield, Illinois, and spends his non-writing time as an abstract painter, independent publisher, and occasional filmmaker.
