Faced with the development of the artificial intelligence-based technologies, do we need to modify our civil liability (tort liability) system or compensation system in general? Although an AI system has several features compared to other technologies, it is its “autonomy” which poses the essential question whether a harm caused by its output can be attributed to a human. The aim of this paper is to overview and to analyze the current theoretical situation regarding this topic in Japan. Just like in other countries, the adaptability of the existing civil liability regime to the accidents caused by autonomous vehicles is considered as a question to be solved as immediately as possible in Japan. Although the liability regime established by the Act on Securing Compensation for Automobile Accidents in 1955 will work acceptably well for the SAE Level 4 (High Driving Automation) vehicles, there are worries that it won’t for the SAE Level 5 (Full Driving Automation) vehicles because of the eventual disappearance of the person responsible according to this regime. This is why the Japanese government and law scholars have started to discuss the alternative regimes aiming to reinforce the civil liability of manufactures of autonomous vehicles such as renewing the compulsory liability insurance, elaborating an independent liability regime for autonomous vehicles, creating a compensation fund, etc. How about AI systems in general? The modernization of the common civil liability regimes will be inevitable: the traditional fault liability regime should be reinforced by introducing the general measures of the presumption of fault and/or causality; the existing product liability regime established by the Product Liability Act in 1994 should be modified to be adapted to the digitalization. However, these solutions might be nothing more than stopgap measures, considering the inherent limits of these regimes (necessity to identify a “misconduct” of a human using an AI system, difficulty in defining a “defect” of an AI system, etc.). Three new regimes are to be further discussed: no fault liability focusing on the uncontrollability of autonomous AI systems, vicarious liability regarding AI systems as auxiliaries of humans, and compensation fund intended to indemnify victims of the outputs of AI systems. Each regime poses a lot of theoretical questions to which the current Japanese civil liability doctrine is not yet ready to answer satisfactorily.
Read full abstract