When physicians and AI work together, who is accountable? (www.nature.com)

A recent discussion highlights the implications of integrating advanced AI tools in healthcare, particularly in terms of accountability. As AI technology evolves, future systems are expected to make independent decisions regarding diagnoses and treatment plans, potentially reducing clinician oversight. This shift raises significant concerns about legal liability, as current frameworks do not adequately address scenarios where both AI and human practitioners contribute to patient care. The opaque nature of advanced algorithms complicates the attribution of responsibility when patients are harmed, creating "liability gaps" that could deter hospitals from adopting such innovative tools. To address these accountability challenges, experts propose a structured liability framework that categorizes medical AI according to three properties: autonomy, automation, and operational scope. They suggest defining seven distinct levels of AI capability, from minimally autonomous systems that solely provide information to complex AI-driven tools capable of managing extensive treatment protocols. This structured approach, reminiscent of regulatory classifications in other high-stakes fields like aviation, aims to clarify responsibilities among healthcare providers, regulators, and AI developers, ultimately facilitating safer integration of AI into clinical practice.