Does the EU Need an Artificial Intelligence Liability Directive? Insights from the Economics of Federalism

Confronted with increasing risks of harm created by various AI systems, the EU authorities have proposed an AI liability directive (AILD) with an intention to address the procedural obstacles suffered by a claimant in a non-contractual fault-based claim and probably in the future, to harmonise the liability and damage rules for the harm caused by AI systems. After vigorous debate, the Commission decided to withdraw the proposal in February 2025. This article provides an explanation of this withdrawal by arguing that the idea of establishing liability rules for AI-related harm at the European level cannot be well justified by the insights from the economics of federalism. Compared with the centralised approach, the already existing decentralised pattern may better satisfy the different local preferences.