Research on Liability Allocation Mechanisms for Harm Caused by Medical Artificial Intelligence under the Civil Law Framework

Abstract

Medical artificial intelligence distributes clinical decision-making among doctors, hospitals, software companies, manufacturers, system integrators, and data and infrastructure providers. When harm occurs, this distributed responsibility complicates traditional civil-law determinations of fault, product defect, causation, disclosure, and proof. Drawing on foreign legal frameworks and China's existing civil‐liability rules, this paper examines the principles of civil liability and product safety legislation in China.It believes that neither the exclusive liability of clinicians nor general enterprise liability is suitable for handling the risks of medical AI.Patients may pursue claims under professional negligence, organisational fault, breach of contract, or product liability.Liability should then be allocated according to operational control, preventability, risk creation, breach of professional or technical duty, evidentiary failure, and causal contribution. A structured formula should guide apportionment, without being fully automated. Rebuttable evidentiary presumptions, mandatory versioned logging, proportionate disclosure, independent auditing, post-deployment monitoring, risk-sensitive insurance, etc., can all reduce information asymmetry and protect legitimate rights and interests. With good calibration, joint and several liability can provide compensation and accountability for harm without labelling every unfortunate event as misconduct and thus prevent stifling good innovation.