AMA Sets Clinical AI Liability Framework

The American Medical Association (AMA) on Thursday outlined how physicians, healthcare organizations and AI developers should share responsibility when clinical AI contributes to harm. Its Center for Digital Health and AI said existing liability frameworks leave unanswered questions and could expose physicians to risks they cannot reasonably prevent while making it difficult for patients to seek remedies.

AMA said current rules are unclear about when product liability applies to AI, how developer contracts and disclaimers can separate responsibility from control of risks, and whether healthcare organizations must conduct local testing or monitor outcomes. Its proposed framework assigns responsibility according to each party’s ability to prevent harm. Multiple parties may be liable, and responsibility is shared where control overlaps, but one party’s obligations do not disappear because another also has responsibility.

For physicians, the framework ties liability to clinical judgment they can meaningfully exercise, considering a system’s intended use, known limitations and information available at the point of care. Physician oversight cannot replace adequate evidence, appropriate system design or conditions that make clinical review meaningful. Doctors should be able to decline an AI system without organizational penalty when its clinical validation, usefulness or place in the standard of care remains uncertain.

AMA President Willie Underwood III said patients need meaningful ways to seek recourse when clinical AI causes harm, while physicians need confidence that responsibility reflects what they can reasonably know and control. He said developers, healthcare organizations and physicians each have responsibilities as AI takes on more clinical functions. The liability framework expands one part of a broader clinical AI governance framework that AMA and its Center unveiled earlier this month. Those principles will guide AMA’s work with providers and stakeholders on issues including licensure and regulation. AMA CEO John Whyte said the framework will guide work with physicians, patients, developers and policymakers on evidence, oversight that changes with risk, and accountability.

The American Medical Association (AMA) on Thursday outlined how physicians, healthcare organizations and AI developers should share responsibility when clinical AI contributes to harm. Its Center for Digital Health and AI said existing liability frameworks leave unanswered questions and could expose physicians to risks they cannot reasonably prevent while making it difficult for patients to seek remedies.

AMA said current rules are unclear about when product liability applies to AI, how developer contracts and disclaimers can separate responsibility from control of risks, and whether healthcare organizations must conduct local testing or monitor outcomes. Its proposed framework assigns responsibility according to each party’s ability to prevent harm. Multiple parties may be liable, and responsibility is shared where control overlaps, but one party’s obligations do not disappear because another also has responsibility.

For physicians, the framework ties liability to clinical judgment they can meaningfully exercise, considering a system’s intended use, known limitations and information available at the point of care. Physician oversight cannot replace adequate evidence, appropriate system design or conditions that make clinical review meaningful. Doctors should be able to decline an AI system without organizational penalty when its clinical validation, usefulness or place in the standard of care remains uncertain.

AMA President Willie Underwood III said patients need meaningful ways to seek recourse when clinical AI causes harm, while physicians need confidence that responsibility reflects what they can reasonably know and control. He said developers, healthcare organizations and physicians each have responsibilities as AI takes on more clinical functions. The liability framework expands one part of a broader clinical AI governance framework that AMA and its Center unveiled earlier this month. Those principles will guide AMA’s work with providers and stakeholders on issues including licensure and regulation. AMA CEO John Whyte said the framework will guide work with physicians, patients, developers and policymakers on evidence, oversight that changes with risk, and accountability.