The Veterans Health Administration operates two general-purpose generative AI chat tools — VA GPT and Microsoft 365 Copilot Chat — for work that may involve patient information, drawing a combined 15,832 active users across two internal Microsoft Teams communities over a 90-day window from October 9, 2025 to January 6, 2026, in a VA OIG national review published on June 11, 2026.
In Report 26-00182-140, the OIG directed three recommendations to the Under Secretary for Health covering permissible clinical uses for general-purpose chat tools and their risk mitigation, evaluation of whether high-impact safeguards applied to Ambient AI Scribe should also apply to these tools, and integration of AI risk monitoring into existing patient safety programs with staff training — all targeted for completion by April 2027.
The underlying tension is one of governance classification. VA identified Ambient AI Scribe, a targeted clinical documentation tool, as high-impact, triggering pre-deployment testing and human-oversight requirements, but did not classify VA GPT or Copilot Chat as high-impact despite their use in clinical workflows. VA leaders described the chat tools as analogous to search engines and emphasized user-level responsibility — a posture the OIG found inconsistent with how the tools were actually being used.
The OIG initiated the national review on October 14, 2025, issued a preliminary result advisory memorandum flagging a patient safety risk on January 15, 2026, conducted its last interview on February 2, 2026, and completed evidence analysis on February 5, 2026 before publishing the final report.
VA's 2025 AI coordination ran through two bodies without a formalized governance relationship: VHA's National AI Institute (NAII) and VA's Office of Information and Technology's Chief AI Officer (CAIO) team. At the time of the review, neither chat tool had web search enabled, so their knowledge base was not current — Copilot Chat web search has since been enabled.
The two AI-focused Microsoft Teams communities drove substantial staff engagement. Over that same 90-day window, CAIO's AI Teams site had 10,997 active users, 1,303 posts and replies, and 1,933 reactions, while NAII's site had 4,835 active users, 585 posts and replies, and 1,043 reactions. In a sample of clinical prompts shared by staff, drafting of clinical notes and summarization of patient care were the most common use cases.
VA's action plan places the AI in Healthcare Workgroup, co-chaired by the VHA National AI Institute and VA Chief AI Office, in charge of developing clinical-use guidance and quality assurance regimes with pre-deployment assessment, ongoing monitoring, use-case education, and transparency requirements. The National Center for Patient Safety is leading parallel work to update the Joint Patient Safety Reporting system so AI-generated documentation can be tagged, traced, and investigated — closing the gap the OIG identified.