VA’s AI chatbots not designated high-impact, despite clinical use, watchdog says
VA’s AI chatbots not designated high-impact, despite clinical use, watchdog says
Publish Date: 2026-06-12 13:44:00
Source Domain: www.nextgov.com
The Office of Inspector General (OIG) of the Department of Veterans Affairs (VA) issued a report highlighting the VA’s failure to classify its generative artificial intelligence (AI) chatbots as high-impact use cases, despite clinicians utilizing them for patient documentation. This lack of classification is raising concerns about patient safety and oversight, as the chatbots, VA GPT and Microsoft 365 Copilot Chat, are not tailored for clinical use and lack central curation or evaluation pertinent to clinical decision-making. Unlike the ambient AI scribe, which follows rigorous safety requirements, the chatbots lack appropriate safeguards, risking errors that could affect patient diagnosis and treatment. The OIG made three recommendations: overseeing generative AI chat tools, ensuring they are evaluated as high-impact, and integrating monitoring into patient safety programs. The VA agreed in principle with these oversight recommendations and concurred with additional measures.
Key Points:
– VA chatbots like VA GPT and Microsoft 365 Copilot Chat are used in clinical settings despite not classified as high-impact.
– Oversight and safeguards are insufficient, creating risks for patient safety and error identification.
– The OIG identified clinical prompts shared in an AI-focused channel, highlighting the role of prompt techniques in output errors.
– OIG recommended enhanced oversight, assessing chatbots as high-impact, and integrating AI risk monitoring into patient safety programs.
– VA has increasingly adopted AI tools, with 215 use cases classified as high-impact out of 367 total.
