We study the market for AI systems that are used to help to diagnose and treat diseases, reducing the risk of medical error. Based on a two-firm vertical product differentiation model, we examine how, in the event of patient harm, the amount of the compensation payment, and the division of this compensation between physicians and AI system producers affects both price competition between firms, and the accuracy (quality) of AI systems. One producer sells products with the best-available accuracy. The other sells a system with strictly lower accuracy at a lower price. Specifically, we show that both producers enjoy a positive market share, so long as some patients are diagnosed by physicians who do not use an AI system. Any transfer in compensation payment from the physician to the AI producer in the case of a diagnostic error will be passed on in full to the physician via the price of the AI system. The quality of the AI diagnosis system is independent of how any compensation payment to the patient is divided between physicians and producers. However, the magnitude of the compensation payment matters. An increase in compensation increases demand for both AI systems. In addition, the higher the compensation paid to the harmed patient, the higher the quality of the low-quality AI system. As the other firm continues to offer the highest accuracy level, any increase in compensation will decrease vertical differentiation, thereby increasing price competition between firms.