The American Medical Association (AMA) confirmed in a recent guidance report that responsibility for errors resulting from AI tools in the medical sector does not rest solely on physicians, but must be fairly distributed among the developers of these technologies and the health institutions that adopt them.
Distribution of Responsibilities in an AI Environment
According to the new framework, each party has a specific role in mitigating risks:
- Developers: They bear the primary responsibility for designing safe and accurate tools, and providing transparent information regarding intended use, performance, and technical limitations, rather than merely relying on legal disclaimers.
- Health Institutions (Operators): Hospitals and health systems are responsible for selecting appropriate tools, evaluating them before deployment, providing necessary training to physicians, and monitoring model performance to ensure accuracy does not decline over time.
- Physicians: They remain responsible for exercising their clinical judgment, including reviewing AI outputs and rejecting recommendations when necessary, but they should not serve as a "safety net" for technical errors they cannot control or identify.
Why Does This News Matter?
This step comes at a time when AI adoption is increasing to support clinical decision-making and reduce administrative burdens. However, a survey conducted by the American Medical Association in 2026 showed that one-third of physicians consider the clarity of legal responsibility frameworks to be the primary driver for adopting these technologies. The survey also revealed a significant training gap, as one-quarter of physicians had received no training on using AI tools, while 92% expressed a desire for additional training.
What Changes in Practice?
The framework aims to establish clear standards that provide those affected by AI errors with legal pathways to obtain compensation from the party actually responsible. The Association argues that this definition of responsibilities not only protects physicians from unfair litigation but also enhances patient safety by requiring each party to implement best practices within their scope of expertise. Furthermore, in its comments directed to the U.S. Food and Drug Administration (FDA), the Association called for developers to be held responsible for monitoring medical devices that rely on generative AI, to avoid overburdening healthcare providers, especially in small and rural institutions.