Evidence and Liability in High-Risk AI under EU law

Abstract

This thesis examines the evidentiary challenges arising in civil liability for high-risk AI systems under EU law. It focuses on the interaction between the Product Liability Directive (PLD), the Artificial Intelligence Act (AIA), and the Digital Services Act (DSA), analyzing how these frameworks shape the availability, accessibility, and use of information in liability claims. The study argues that while the PLD provides the core liability regime, its effectiveness is limited by structural difficulties in proving defect and causation in complex AI systems. In this context, the AIA and DSA are conceptualized as regulatory frameworks that contribute to the creation and organization of information relevant for evidentiary purposes. The thesis concludes that the EU framework can partly address proof-related challenges, but only if regulatory documentation and platform-based transparency mechanisms can be effectively utilized within national civil procedures.