Civil liability regime arising from the use of artificial intelligence in arbitration

Abstract

This article examines the reconfiguration of the civil liability regime governing the use and integration of artificial intelligence (AI) in international arbitration. Based on a dogmatic distinction between AI used as a professional support tool, on the one hand, and autonomous AI systems (such as the so-called “AI arbitrator”), on the other hand, this article evaluates the applicability of both fault-based (subjective) and strict (objective) liability regimes. The study analyzes the core elements of civil liability (generating act or wrong, compensable damage, and causal link) within the conduct of the main arbitration actors, particularly arbitrators and counsel, with particular emphasis on compliance with soft law and hard law frameworks, duties of transparency, confidentiality, and the fundamental principle prohibiting the delegation of adjudicative functions. The article concludes that while the assistive use of AI is governed by a fault-based liability centered on human oversight and professional diligence of the person using it, autonomous AI requires strict or hybrid liability models. Finally, it cautions against the risks of setting aside or refusing recognition of arbitral awards due to due process and public policy violations, wondering about a possible reinterpretation of traditional procedural principles such as iura novit curia.