Optimal Liability Rules for Combined Human-AI Health Care Decisions

The integration of AI for healthcare redefines medical liability, transforming decision-making into a collaborative process between technology and its user. When a harm is caused, both AI users and manufacturers may be responsible. The judicial system has yet to address claims of this nature. We develop a model with bilateral care to analyze which liability rules lead to socially efficient investment in care by AI producers and users. Both parties may be subject to strict liability, negligence rules, or hybrid regimes-where one agent operates under strict liability while the other is subject to fault-based liability. For each regime, we examine the role of the compensation-sharing scheme between users and producers. The European Parliament's latest AI Liability Directive for the medical field supports a strict liability regime for AI producers and a fault-based liability regime for AI users. Our findings confirm that this framework achieves social efficiency in healthcare. While a new regulatory framework is not strictly necessary, we also identify an alternative socially efficient regime in which the physician alone assumes full medical liability.