This issue presents a multi-layered challenge, as the resulting harm may stem from model design, training data, hospital deployment, system updates, or the physician's decision. Employing a descriptive-analytical method and a comparative jurisprudential-legal approach, this study examines the grounds for civil liability regarding these systems under Iranian law and compares them with early, later, and contemporary jurisprudential views, French law, and international instruments. The analysis reveals that, despite exhibiting quasi-agency in behavior, the system lacks legal capacity and the ability to directly bear liability (*dhaman*); therefore, the roles of the manufacturer, supplier, medical facility, and physician must be evaluated based on customary attribution, risk creation, control capabilities, and access to information. Legal principles such as direct destruction (*itlaf*), causation (*tasbib*), the "stronger cause" doctrine, deception (*ghurur*), the "no-harm" rule (*la-darar*), and physician liability offer a framework for analyzing this chain of events; however, relying solely on proving individual fault is insufficient given the technical complexity and information asymmetry involved. The comparative study suggests that—while maintaining the physician's professional liability—it is possible to establish multi-layered liability, obligations regarding documentation and evidence disclosure, mandatory insurance, and compensation funds to address product defects and high-risk systems. The article concludes that a coherent response lies not in attributing legal personality to artificial intelligence, but in distributing liability among the actors who possess a more realistic capacity for prevention, control, and risk distribution.