Responsible Artificial Intelligence for Veterans Act of 2026
- Sponsor
- Rep. Murphy, Gregory F. [R-NC-3]
- Committees
- Veterans' Affairs Committee (primary)
- Last action
- Jul 29, 2026
Bottom line
This bill mandates an independent, comprehensive audit of AI systems within the VHA to ensure their safety, effectiveness, and ethical deployment for veteran care, funded through existing VA appropriations.
What it actually does
This bill directs the Secretary of Veterans Affairs to enter into an agreement with a federally funded research and development center (FFRDC) to conduct an independent evaluation of artificial intelligence systems deployed, in pilot, or in active development for clinical use within the Veterans Health Administration (VHA). The evaluation must prioritize at least five high-risk or widely deployed AI systems, assessing their integration, governance, performance, safety, veteran-centric design, cost, and transparency. The FFRDC will report its findings and recommendations to Congress, and the VA must then submit a corrective action plan.
Proponents argue
Supporters argue this bill is crucial for ensuring that AI technologies deployed within the VHA are safe, effective, and free from bias, protecting veterans from potential harm while leveraging AI's benefits. They emphasize the importance of independent oversight to build trust and accountability in critical healthcare systems, particularly given the rapid advancement of AI.
Opponents contend
Potential opponents might argue that mandating an FFRDC evaluation could be redundant with existing VA internal oversight mechanisms or that it could unnecessarily delay the deployment of beneficial AI systems due to bureaucratic processes and external review. Concerns might also be raised that requiring the use of existing funds could divert resources from other critical VA initiatives.
The bill is short and clearly written, allowing for a quick and thorough understanding of its provisions within a typical legislative review period.
Section 2(a)
Mandate for Independent AI Evaluation of VHA Systems
This provision requires the Secretary of Veterans Affairs to enter into an agreement with a federally funded research and development center (FFRDC) within 90 days of the bill's enactment. The FFRDC will conduct an independent evaluation of artificial intelligence systems currently deployed, in pilot, or under active development for clinical use within the Veterans Health Administration. The evaluation must prioritize at least five AI systems that are widely deployed or pose significant clinical, operational, or patient safety risks.
Supporters argue
Supporters argue this ensures an unbiased, expert assessment of VHA's AI systems, identifying potential flaws or risks that internal reviews might miss. This independent layer of scrutiny is vital for patient safety and system integrity, fostering trust in AI technologies.
Critics contend
Some might argue that the VA already has internal mechanisms for evaluating technology, and an external FFRDC contract could add unnecessary bureaucracy or cost, potentially duplicating existing efforts and slowing down innovation.
Tradeoffs
This provision balances the need for independent oversight and accountability in critical healthcare technology with potential concerns about administrative burden or perceived redundancy with existing internal VA processes.
Section 2(b)
Comprehensive Evaluation Criteria for AI Systems
This section details the specific aspects an independent evaluation must address for each AI system. These criteria include integration and operational readiness, governance and accountability, performance and model integrity (including bias mitigation), safety and human oversight, veteran-centric design, cost and resource stewardship, transparency and documentation, and scalability and overall effect (including environmental impact). This ensures a holistic review of AI systems.
Section 2(c)
Risk-Based Selection Criteria for AI Evaluation
This provision specifies that the FFRDC should prioritize AI systems for evaluation that influence critical decisions (diagnosis, treatment, triage, eligibility), are widely deployed or impact a significant veteran population, use advanced AI techniques (predictive analytics, large language models, imaging analysis), or present material cybersecurity, privacy, or patient-safety risks. This ensures the evaluation focuses on the most impactful and potentially vulnerable systems.
Section 2(d)
VA Information Sharing Requirements for Evaluation
The Secretary of Veterans Affairs must ensure the FFRDC has timely access to all relevant VA information needed for the evaluation. This includes contracts, technical documentation, validation studies, model cards, performance data, audit logs, incident reports, corrective action plans, and policies on AI governance, data use, and risk management. This comprehensive access is designed to enable the FFRDC to conduct a thorough and informed review.
Section 2(e)
Reporting Requirements and Accountability Framework
This provision outlines a series of reporting requirements. The FFRDC must submit a report within one year of the agreement, detailing findings, identifying risks, recommending corrective actions, and flagging systemic gaps. This report goes to the Secretary and relevant Congressional committees. Within 120 days of receiving this report, the Secretary must submit a response and a corrective action plan to Congress, including specific actions, assigned responsibilities, milestones, deadlines, and justifications for unmitigated risks. The Comptroller General will then review both the evaluation and the VA's plan.
Section 2(f)
Funding Through Existing Appropriations
This provision explicitly states that the Secretary shall carry out the evaluation using amounts already authorized and appropriated to the Department of Veterans Affairs. It clarifies that no additional amounts are authorized to be appropriated for this section, meaning the cost must be absorbed within the VA's existing budget.
Defines "artificial intelligence" by cross-referencing section 5002(3) of the National Artificial Intelligence Initiative Act of 2020 (division E of Public Law 116-283; 134 Stat. 4523; 15 U.S.C. 9401 note).
Section 2(g)
Why it matters:This is standard legislative practice to ensure consistency with existing federal definitions and avoid redundancy across different statutes.
Case for: Using an established federal definition ensures consistency across government initiatives and leverages existing legal frameworks for AI, providing clarity and avoiding conflicting interpretations.
Case against: Relying on external definitions requires readers to consult other statutes to fully understand the scope of the bill, potentially obscuring its reach for those unfamiliar with the referenced law.
Estimated impact: Ensures a consistent and broad understanding of what constitutes "artificial intelligence" for the purpose of this evaluation, encompassing various technologies from machine learning to expert systems, thereby defining the scope of the entire audit.