AMA Releases New Liability Framework for Clinical AI

The framework aims to assign accountability for patient harm among doctors, developers, and healthcare institutions.

Updated on Oct. 30, 2026 in Artificial Intelligence

Bold flat-color editorial illustration showing a clinical sensor balanced against an abstract structural block, representing medical accountability policy. — The American Medical Association released a new liability framework outlining how accountability for clinical AI-related harms is shared among developers, hospitals, and physicians. AI Illustration. Upload story photo >

Live Poll

Should physicians or AI developers be held primarily accountable for errors in clinical AI software?

The American Medical Association has released a new liability framework designed to assign accountability for clinical AI-related harms. This policy outlines how responsibility should be distributed among developers, healthcare institutions, and physicians.

Why it matters

Current liability structures lack mechanisms to address risks unique to medical AI, leaving patients without clear avenues for legal recourse. This framework attempts to bridge that gap by aligning accountability with the party best positioned to prevent harm.

The framework establishes a liability model that delegates responsibility to the party with the greatest capacity to prevent harm. It formally defines roles for developers, organizations, and clinicians, replacing a landscape where professional liability standards for AI remained undefined.

The players

American Medical Association

The national professional association representing physicians in the United States that sets ethical and clinical standards.

The details

The framework utilizes a control-based approach to determine liability, placing the burden on the entity most capable of managing technical and clinical risks. Physicians are required to maintain final authority over patient care, applying clinical judgment based on the specific intended use and documented limitations of the AI tools involved. This ensures that while developers and institutions face accountability for system failures, the clinical standard of care remains centered on human decision-making.

October 2026: The AMA released broader governance principles for clinical AI.
October 29, 2026: The AMA published the specific liability framework.

The Tech Race

This framework follows the AMA's October 2026 clinical AI governance principles, marking a shift from high-level guidance to actionable liability standards. It positions the organization as a central architect in the evolving regulatory environment for medical software.

Healthcare providers will see this framework inform future medical licensure and institutional practice standards across the United States. While not a law itself, it sets the professional expectations that will guide how physicians integrate and oversee AI tools in patient workflows.

The takeaway

The AMA is now prioritizing the integration of this liability model into regulatory and licensure discussions with external stakeholders. Physicians and health administrators should monitor upcoming state-level adoption of these principles as the primary signal of shifting legal expectations.