AI Liability Along the Value Chain: Lessons from the Liability of Suppliers of Components in Product Liability Law

Abstract Though it is common to think of AI systems as a singular tool, AI systems are often developed and deployed in a value chain that involves numerous components and actors. The AI life cycle has at least three stages: programming, fine-tuning, and implementation. Different actors can participate in these stages or sell and supply key components, such as pre-labeled data. Consequently, when harm occurs and an AI system is involved, allocating liability among these different actors and stages poses difficult questions. This Article begins to address this question by drawing on existing tort doctrine on the liability of suppliers of raw materials and components in product liability law. In doing so, it identifies principles that can inform courts, regulators, and technologists on whether and when the different participants in the AI value chain should be held liable for harm involving an AI system. It argues that several factors should be considered when facing AI liability cases. These include the relative expertise of the different actors, the moment in the AI lifecycle where it may be most efficient to adopt safety measures, the context of the AI system’s deployment, whether AI component suppliers should foresee specific uses and risks, and the level of participation of each AI actor in the outcome that causes harm.