Law and Economics of Artificial Intelligence: Optimal Liability Rules for Accident Losses Caused by Fully Autonomous Vehicles

We examine the optimal liability rule in accidents involving fully autonomous vehicles. In cases where enforcing due activity is not feasible, it is socially optimal to apply the strict liability rule to the human operator determining the activity level and to apply the negligence rule to the manufacturer and the victim who select care levels under contributory or comparative negligence in the unilateral activity case. Under the joint and several liability rule, both the manufacturer and the victim exercise due care, contingent on regulating the manufacturer’s liability share sufficiently high, and the human operator assuming the remaining risks chooses the socially optimal activity level maximizing the social net benefit. Conversely, if due activity enforcement is possible, an alternative liability rule proves optimal. Under this rule, the human operator engages in efficient activity to comply with the activity standard, the manufacturer exercises efficient care to meet the care standard, and the victim assumes residual liability so as to be induced to take efficient care. Notably, this liability rule achieves the social optimum, even in bilateral activity cases where both the human operator and the victim engage in activity. Our results diverge from previous findings suggesting that achieving the social optimum involves using public sanctions, such as paying a fine to the state.