⚡NIMITZ NEWS FLASH⚡

“Advancing VA Care Through Artificial Intelligence”

House Veterans Affairs Committee, Technology Modernization Subcommittee Hearing

September 15, 2025 (recording here)

HEARING INFORMATION

Witnesses & Written Testimony (linked):

• Mr. Charles Worthington: Chief Technology Officer & Chief Artificial Intelligence Officer, Office of Information & Technology, U.S. Department of Veterans Affairs

• Dr. Evan Carey: Acting Director, National Artificial Intelligence Institute, Veteran Health Administration, U.S. Department of Veterans Affairs

• Dr. Mohammad Ghassemi: Assistant Professor, Michigan State University

• Mr. Sid Ghatak: Chief Technical Advisor, National Artificial Intelligence Association

• Ms. Carol Harris: Director, Information Technology and Cybersecurity Issues, U.S. Government Accountability Office

Keywords mentioned:

• Artificial intelligence (AI), VA healthcare, use cases, data privacy, governance, workforce, security, implementation, transparency, bias, automation, research, modernization

IN THEIR WORDS

OPENING STATEMENTS FROM THE SUBCOMMITTEE

• Chairman Tom Barrett opened the hearing by stressing the importance of ensuring that the VA’s technology is efficient and reliable to support, rather than hinder, veterans’ care. He said artificial intelligence (AI) might sound futuristic or intimidating to some but is increasingly relevant to improving VA services. He highlighted the Subcommittee’s duty to ensure the VA uses technology effectively to close gaps in care and service.

• Ranking Member Nikki Budzinski cautioned that while AI holds promise, the VA continues to struggle with basic IT infrastructure and modernization. She expressed excitement about AI’s potential in clinical care, such as cancer detection and easing clinician burnout, but argued that success required resources, transparency, and strong workforce investments. She criticized the VA’s lack of a confirmed Chief Information Officer (CIO), warning that AI integration risked happening in silos without leadership or a coherent strategy. She also raised concerns about data privacy, cybersecurity breaches, workforce cuts, and governance, stressing the need for safeguards before AI could be responsibly deployed.

SUMMARY OF KEY POINTS

• Mr. Charles Worthington outlined the VA’s strategic vision to be a leader in AI, focused on delivering faster services, higher quality care, and cost-effective operations. He described five priorities: expanding AI access, reimagining workflows, investing in data infrastructure, cultivating an AI-ready workforce, and ensuring transparent governance. He cited examples such as the REACH VET program, AI-assisted colonoscopy, and tools saving staff hours weekly, while stressing rigorous standards for privacy and security. Mr. Worthington acknowledged challenges in integration, talent recruitment, and costs, urging congressional support to sustain progress.

• Mr. Sid Ghatak drew on decades of AI experience in government, finance, and healthcare to argue that AI extends far beyond large language models (LLMs) like ChatGPT. He cautioned that over-reliance on LLMs risked overlooking other powerful AI technologies designed for specialized tasks, some of which could run on small devices like phones. Using the steam engine as an analogy, he said revolutionary technologies evolve over time and require collaboration between the government and the private sector to mature safely. He believed AI was at the start of a revolution comparable to the Industrial Revolution and that the VA’s mission would be advanced by diverse AI innovations.

• Dr. Mohammad Ghassemi described AI’s role in health care as falling into three categories: automation, augmentation, and insights. He gave examples such as AI scribes reducing paperwork, emergency tools triaging patients faster, and monitoring systems detecting early health declines. He explained that AI could streamline operations by reducing missed appointments, preventing gaps in care, and ensuring critical findings were followed up on. He concluded that while AI is not a silver bullet, disciplined pilots, clear metrics, and safeguards could make it a valuable tool to return time to patients and support clinicians.

• Ms. Carol Harris stated that AI could greatly improve VA operations but carried oversight risks given the VA’s longstanding IT management challenges. She reported that the VA’s AI use cases had increased from 40 in 2023 to 229 in 2024, with many considered “high impact” to individual safety and rights. She highlighted challenges in complying with federal AI policy, resource and funding shortages, and building an AI-ready workforce. She urged the VA to implement the Government Accountability Office’s (GAO’s) 26 open recommendations on IT management and to adopt GAO’s framework for responsible AI use, organized around governance, data, performance, and monitoring.

• Chairman Barrett asked how the VA ensured vendors did not use veteran data for secondary purposes. Mr. Worthington explained that all AI systems must meet stringent VA security and privacy standards, receive an authority to operate, and require vendors to attest to compliance.

• Chairman Barrett then asked whether LLMs could be used to predict risk factors for veterans presenting with certain conditions. Dr. Evan Carey explained that providers could use AI-based decision support tools to follow recommended procedures and identify guidelines.

• The Chairman asked whether data from the burn pit registry and similar sources could benefit broader medical research. Mr. Worthington confirmed that the VA had a long tradition of advancing general healthcare research and cited Dr. Carey’s recent AI-driven studies. Chairman Barrett pressed on whether vendors could profit from VA data without benefiting veterans, and Mr. Worthington agreed that safeguards were needed to ensure mutual benefit.

• Ranking Member Budzinski asked how the VA was educating veterans about opt-in AI use cases. Mr. Worthington explained that the VA published its AI strategy and use case inventory and actively sought veteran feedback to maintain trust.

• Ranking Member Budzinski questioned Mr. Worthington about reports of specific employees possibly accessing VA data for AI operations. Mr. Worthington confirmed knowing several but said he was unaware of any improper data use or duplication. He committed that, to his knowledge, no veterans’ data had been removed.

• The Ranking Member also asked if the VA agreed with calls to recruit more tech talent. Mr. Worthington strongly agreed, citing partnerships with the U.S. Digital Corps and the Presidential Innovation Fellows Program.

• Ranking Member Budzinski asked Ms. Harris how a skilled workforce supported AI scalability and data protection. Ms. Harris answered that expertise was critical not only to maximize benefits but also to address cybersecurity, bias, and intellectual property risks.

• Rep. Morgan Luttrell raised concerns about fragmentation across 172 VA facilities and questioned the comprehensiveness of the AI inventory. Mr. Worthington acknowledged challenges but said the VA was deploying enterprise templates and AI-assisted medical devices.

• Rep. Luttrell asked whether AI deployment should wait until systems were standardized. Mr. Worthington claimed that the VA could not wait, since AI was embedded in private-sector tools, and advocated for standard templates.

• Rep. Sheila Cherfilus-McCormick asked if the VA could truly standardize AI, given past electronic health record (EHR) failures. Mr. Worthington described structured pilots and enterprise approaches. She pressed on preventing bias, and Mr. Worthington cited OMB requirements for pre-deployment testing and monitoring.

• When Rep. Cherfilus-McCormick asked, Ms. Harris emphasized that 72% of VHA’s AI use cases were “high impact” and urged filling more privacy officer positions. Dr. Ghassemi noted research successes in bias detection, saying de-biasing was possible but dataset-specific.

• Ranking Member Budzinski highlighted Suicide Prevention Month and asked how the VA would build on REACH VET. Dr. Carey explained that the model was recently updated and incorporated into a broader strategy, with clinicians always in control. The Ranking Member secured confirmation that AI would not replace frontline responders.

• Ranking Member Budzinski also cited an Office of Inspector General (OIG) finding that an automated VBA tool had misassigned PACT Act claim dates, leading to $7 billion in improper payments. She asked how accuracy would be measured. Mr. Worthington replied that each use case had an owner and was subject to continuous monitoring.

• Rep. Luttrell asked whether consolidating all veterans’ data in a secure enclave would reduce theft risks. Mr. Worthington agreed that it was helpful and cited REACH VET’s reliance on a central warehouse. Mr. Ghatak stressed the need for enterprise-level leadership and investment in data alignment.

• Chairman Barrett asked about AI guardrails. Mr. Ghatak responded that models lacked moral judgment and industry relied too heavily on correlation, urging greater investment in causal methods and data quality.

• Ranking Member Budzinski reiterated the need for transparency on six employees’ data access, referencing a letter from Full Committee Ranking Member Mark Takano.

• Chairman Barrett illustrated AI’s potential to improve care and save lives but insisted it must not replace clinicians, and that trust depended on safeguarding data and transparency.

SPECIAL TOPICS

JOIN THE NIMITZ NETWORK!

Enjoying our updates? Don’t keep it to yourself — forward this email to friends or colleagues who’d love to stay informed. They can subscribe here to become part of our growing community.