Building on its recent guidance for organizations’ use and governance of clinical AI, the American Medical Association on Thursday gave a deeper breakdown of how it believes different parties should be responsible for decisions, actions and failures involving the technology.
The latest framework from AMA’s Center for Digital Health and AI addresses liability issues the group said are currently unanswered by existing liability frameworks. Those structures, it said, could leave physicians facing "liability for risks they cannot reasonably prevent, while patients may struggle to seek remedies when AI causes harm."
Other key issues with the status quo liability frameworks, AMA said, are ambiguity around when product liability applies to AI; how contract terms and disclaimers from developers can distance responsibility from risk control; and whether deploying healthcare organizations hold various responsibilities specific to AI, such as local testing or outcomes monitoring.
AMA said the clinical AI liability framework it’s advancing "recognizes that multiple parties may bear liability for the use of AI when something goes wrong," and then aligns that accountability based on a party’s ability to prevent harm. Responsibility is spread out in areas where control is shared, but it does not become interchangeable, and no party’s obligations disappear because another party also bears some,” AMA wrote in a brief outlining its thinking.
As for the physicians AMA represents, the liability framework calls for doctors to "be responsible for the clinical judgment they can meaningfully exercise, based on the system’s intended use, known limitations and the information available at the point of care."
The group also said that oversight from a physician "cannot substitute for adequate evidence, appropriate system design or the conditions necessary to make clinical review meaningful." Physicians should also retain the authority to decline an unproven AI system, or those "whose clinical validation, usefulness or place in the standard of care remains uncertain," without penalty from their organization, AMA said.
"Patients must have meaningful avenues for recourse when clinical AI causes harm, and physicians must be able to use these technologies with confidence that responsibility reflects what they can reasonably know and control," AMA President Willie Underwood III, M.D., president of the AMA, said in a release on the liability framework. "As AI assumes more clinical functions, developers, healthcare organizations and physicians each have important responsibilities. We need clear expectations that protect patients, support sound clinical judgment, and ensure accountability remains with those best positioned to prevent harm."
The liability framework breaks down in detail one aspect of a broader clinical AI governance framework unveiled by AMA and its Center earlier this month. Those principles and focus areas will guide how AMA works with providers and stakeholders as it tackles issues like licensure and regulation for the fast-moving technology.
"We cannot afford to wait for these questions to be answered through trial and error in patient care," AMA CEO John Whyte, M.D., told Fierce Healthcare of the efforts earlier this month. "The AMA's new framework sets out our ambitious agenda to address these questions. … It will guide our work with physicians, patients, technology developers and policymakers to ensure clinical authority is earned through evidence, how oversight must evolve with risk, and how accountability is maintained."