The Department of Veterans Affairs followed the administration’s push to reduce oversight by providing generative artificial intelligence chat tools to clinical staff without proper oversight or safeguards, a report found Thursday.
An investigation conducted by the OIG from October 2025 to February 2026 found that despite active use of AI tools at the Veterans Health Administration, a lack of proper governance created “risks to patient safety” and limited the ability to monitor errors.
The Veterans Administration is using VA GPT and Copilot Chat to pilot Ambient AI Scribe for drafting clinical notes, which appears to have better governance, according to the report.
“OIG is concerned about VHA’s ability to promote and protect patient safety when these chat tools are used in clinical operations without the same protocols (e.g., feedback loops and pattern detection measures) implemented in Ambient AI Scribe. This could improve clinical use and prevent future patient safety events,” the report states.
OIG also found that VA’s efforts to reduce “unnecessary levels of bureaucratic oversight,” consistent with the Office of Management and Budget’s 2025 AI Memorandum, may have resulted in additional vulnerabilities.
“VA leaders’ focus on reducing barriers to innovation influenced their oversight decisions,” the report said. “The OIG is concerned that this opinion disallows the use of high-impact AI chat tools by VA staff, resulting in inadequate risk management practices such as those specified in the same memo for ‘expected risks.’ [AI’s] use. ‘”
Interviews with AI leaders at the Department of Veterans Affairs found they prioritized quick access to AI chat tools over “potentially unrealistic safeguards,” the report said.
VA AI leaders also did not work effectively with VHA’s National Center for Patient Safety and likened generative AI to a search engine, which the OIG determined was “defective” because generative AI synthesizes information, the report said.
Additionally, the report states that the VA does not have processes or infrastructure in place to ensure patient safety communication and monitoring through AI, including ways to identify records created by AI to retrospectively uncover safety concerns.
OIG recommended that the Department of Veterans Affairs address the governance of generative AI chat tools, assess the high impact of AI chat tools, require safeguards, and integrate monitoring of AI-related risks into existing patient safety programs.
In response to the draft report, the Department of Veterans Affairs agreed with the recommendations, took steps to strengthen communication with AI governance bodies, health agencies, and reporting systems, and provided an action plan to develop clinical AI governance and education.
“VHA greatly appreciates OIG’s assistance in ensuring that all stakeholders come together to support VHA’s vision of providing the highest quality care to all veterans,” said VA Health Secretary John J. Bartram. “Your cooperation will help us realize our commitment to providing excellent health care services for veterans.”
