The American Medical Association (AMA) has introduced a governance framework for agentic artificial intelligence (AI) systems performing clinical functions. The framework, developed by the AMA’s Center for Digital Health and AI, aims to ensure AI improves patient care while maintaining clear lines of clinical responsibility.
This new governance framework outlines principles for clinical authority, performance validation, oversight, and accountability as AI systems take on more consequential roles in healthcare. It is intended to guide the AMA’s work with physicians, patients, technology developers, and policymakers on issues of liability, licensure, and regulatory approaches.
Key Principles of the Framework
A central tenet of the AMA’s framework is that clinical authority for AI systems should be earned through evidence and maintained through demonstrated performance. The principles also state that oversight and authority should adapt based on evidence, system performance, clinical risks, and technological changes. The AMA has identified four key areas for focus: defining an AI system’s clinical role, validating its performance and safeguards, establishing appropriate limits on its authority, and determining accountability for its use, including associated risks and outcomes.
AMA CEO John Whyte, M.D., stated that the growing capabilities of AI systems are raising urgent questions about clinical authority, oversight, and accountability. In a statement, Whyte said the healthcare sector cannot afford to resolve these questions through trial and error in patient care. He added that the framework will guide the AMA’s efforts to establish how clinical authority should be assigned based on evidence, how oversight should adapt to risk, and how accountability should be maintained. Whyte emphasized the need to address these issues before agentic AI systems become more deeply integrated into clinical practice.
AMA’s Expanding AI Governance Efforts
The new framework follows other AMA initiatives addressing AI-related risks in healthcare. In May 2026, the organization released a policy framework to protect physicians from unauthorized AI-generated deepfakes and medical impersonation. This earlier framework included principles for protecting physician identity, prohibiting deceptive medical impersonation, requiring informed consent and transparency, establishing shared responsibility, and providing enforcement mechanisms.
In April 2026, the AMA also urged lawmakers to strengthen safeguards for AI chatbots used in mental health services. Their recommendations included stronger data protection and transparency standards, along with penalties for deceptive practices. The latest framework extends the AMA’s policy efforts to address the clinical authority and accountability questions that emerge as AI systems assume more complex functions in healthcare.
The framework is intended to inform discussions on how healthcare stakeholders should determine the responsibilities and boundaries of AI systems, particularly as they undertake increasingly consequential clinical functions.
