Level 4 autonomous vehicles change the basic assumptions of the traditional legal framework for traffic accident liability. Within a designated Operational Design Domain, the Automated Driving System performs the dynamic driving task without continuous monitoring or immediate intervention by a human driver. The conventional automobile accident liability regime has been structured around driver negligence, operator liability, and compulsory liability insurance for the protection of accident victims. However, in Level 4 autonomous vehicle accidents, the causes of accidents may extend to system-related factors such as sensor malfunction, software errors, defects in perception and decision-making algorithms, communication failures, cybersecurity breaches, over-the-air update defects, and inaccuracies in high-definition maps. As a result, it becomes difficult not only for victims but also for insurers that have provided prompt compensation to identify the precise cause of the accident and the responsible party. The subsequent recourse relationship among manufacturers, system operators, vehicle owners, and insurers may also become unclear.
This study argues that the central task of the liability framework for Level 4 autonomous vehicle accidents is not to abolish operator liability, but to preserve its victim-protection function while refining the internal allocation of responsibility among relevant parties. To this end, the existing structure of prompt compensation followed by recourse under the Compulsory Motor Vehicle Liability Security Act should be maintained as the basic axis of victim protection. At the same time, it should be supplemented in a manner appropriate to Level 4 autonomous driving by redefining grounds for exemption, clarifying the legal basis for recourse claims, imposing duties to preserve and share accident data, strengthening the authority of accident investigation bodies, expanding the treatment of software defects and easing the burden of proof under product liability law, improving the accident reporting and management system under autonomous vehicle legislation, and reorganizing subrogation and recourse structures under insurance law.
Specifically, the insurer’s subsequent recourse should be structured around such factors as whether the Automated Driving System was activated, whether the vehicle operated within the approved Operational Design Domain, whether there was a system defect, whether operational management duties were breached, and whether accident data were properly preserved and submitted. Where multiple responsible parties are involved in the occurrence of an accident, the allocation of liability should be determined by comprehensively considering risk creation, control over the risk, the possibility of risk prevention, and compliance with duties to cooperate in the submission of relevant data. In addition, with respect to software defects and AI decision-making errors, a framework should be established to ease the burden of proof in light of information asymmetry and to provide purpose-limited access to accident data.
In conclusion, the liability framework for Level 4 autonomous vehicle accidents should be redesigned as a dual structure. Externally, it should provide victims with prompt and stable compensation. Internally, it should allocate final responsibility among manufacturers, system operators, vehicle owners, and insurers on the basis of accident data and expert investigation results. Such a structure would enhance predictability for the autonomous vehicle industry while giving practical effect to the victim-protection principle of the Compulsory Motor Vehicle Liability Security Act in the future mobility environment.