Towards a Coherent EU Civil Liability Regime for AI-Caused Harm

Abstract

As artificial intelligence (AI) systems become increasingly embedded in cross-border goods and services, the European Union (EU) has undertaken the ambitious task of harmonising liability and conflict-of-laws frameworks to address AI-caused damage. This article critically examines the interaction between the AI Act (AIA), the revised Product Liability Directive (PLD), and the now-withdrawn AI Liability Directive (AILD). The paper explores whether this regulatory architecture sufficiently provides legal certainty, accountability, and access to redress for victims of AI-related harm, particularly in cases involving multiple jurisdictions. Using doctrinal and comparative legal analysis, the article evaluates the substantive, temporal, and territorial scopes of these instruments and their interface with private international law, especially Article 5 of the Rome II Regulation (Rome II). It argues that while the AIA and revised PLD represent a coordinated ex-ante and ex-postregulatory scheme, gaps remain, particularly regarding immaterial harms, collective injuries, and causal complexity in opaque AI systems. The AILD, though withdrawn, would have played a crucial procedural role in harmonising fault-based liability across Member States, particularly by introducing presumptions of causation and disclosure mechanisms. The article concludes that harmonisation of EU liability frameworks requires not only legislative alignment among the AIA, PLD, and future initiatives, but also targeted reforms to private international law rules. Without such integration, the EU risks fragmented litigation, legal uncertainty, and under-enforcement of AI accountability norms across borders.