AMA Issues Governance Framework for Agentic AI in Healthcare

Summary

The American Medical Association has released a governance framework for using agentic AI in healthcare. It says AI agents that perform clinical functions should be supported by evidence and maintain high-quality performance. The framework asks four questions: what clinical function the agent performs; what evidence shows it can perform that function and what happens if it fails; what the agent may do and under which conditions; and who is responsible for its risks and outcomes. The guidance is aimed at uses such as ambient documentation, diagnostics, and patient monitoring, where agents can complete complex, multi-step tasks with limited human oversight. The AMA warns that inaccurate or biased decisions could harm patients before a clinician can intervene, and emphasizes that physicians should remain central to development and deployment. The framework will also guide the AMA's work with other organizations. Its release comes as adoption expands: a survey of 250 U.S. healthcare leaders found that 28% had deployed agentic AI and 44% were running pilots or proof-of-concept projects, while most expected a transformative effect on clinical workflows. At the same time, 72% said such tools were sometimes implemented without IT approval. More than half identified possible compliance or regulatory violations and unauthorized access to protected health information as major concerns. The article places the framework within a regulatory gap: federal rules remain limited, state approaches vary, and Manatt counted more than 250 healthcare-related AI bills introduced in 47 states in 2025.