The group’s Center for Digital Health and AI laid out its accountability principles in October 2026. The framework itself was released on October 29, 2026.
Formal policy followed shortly after. The AMA adopted it at its Interim Meeting in Orlando on November 10, 2026.
The central idea is that liability should track who actually controls what. Under the new policy, responsibility is divided according to the capabilities and responsibilities of three parties: AI developers, healthcare organizations, and physicians.
In practice, if a clinical error happens because an AI product is defective, and the physician exercised appropriate clinical judgment, responsibility for the harm would fall on the vendor or manufacturer.
The vendor side of the bargain
The AMA is not just asking for a shield. It is also asking AI companies to take on obligations of their own.
Under the framework, AI vendors should maintain adequate product liability insurance. They should also disclose the limitations of their tools and share performance data.
Perhaps most pointed is a provision aimed at contract language. The AMA says vendors should not be allowed to impose terms that unfairly push liability onto healthcare professionals.
The policy also covers pressure from inside the system. The AMA says payers and hospitals should not penalize physicians who decline to use AI tools they consider uncertain.
Patients still need a path to recourse
AMA President Willie Underwood III has stressed that patients must keep avenues for recourse when clinical AI harms them. He has also tied that recourse to confidence, arguing that people need to trust these technologies for them to be used well.
The framework’s answer is to point patients toward the parties that built and deployed the tool when the tool is the problem. That is why the insurance and disclosure requirements for vendors sit alongside the physician protections.
The AMA also frames AI as a complement to physician-led care rather than a substitute for it. The group pushes back on the idea of granting licensure or certification to AI systems, keeping the licensed professional at the center of the care relationship.
Why doctors care about this now
Survey results cited in the AMA’s research show that more than a third of physicians consider clear liability frameworks crucial to building trust in AI and encouraging its adoption in clinical settings.
What this means
For physicians, the policy is an attempt to preserve professional judgment as the deciding factor. Whether a doctor follows the algorithm or overrules it, the AMA wants the legal test to focus on whether that judgment was sound.
The key caveat is that an AMA policy is not law. The association is lobbying for these protections, and legislators would need to act before any of them carry legal force.
