Legal Liability for Harm Caused by Artificial Intelligence Systems

Abstract

This article examines the attribution of liability when harm arises in the development, provision, and deployment of artificial intelligence (AI) systems, with particular emphasis on civil liability for damages. Through doctrinal analysis, engagement with scholarly arguments, and a comparison between Vietnamese and European Union (EU) law, it argues that the autonomy of AI does not, by itself, create a new legal person to bear liability. The central difficulties instead lie in the fragmentation of control over risk and the evidentiary imbalance between injured parties and actors in the AI value chain. Vietnam’s Law on Artificial Intelligence 2025 establishes a compensation obligation for deployers in certain cases involving high-risk AI systems. Nevertheless, it does not fully resolve the problems of proving causation or allocating the ultimate financial burden when the source of harm lies in development or supply. This article proposes a layered framework: an identifiable point of recourse for victims of certain high-risk activities; the continued application of liability grounded in conduct, defects, and control over risk; and statutory rules on evidence preservation, access to evidence, and recourse against responsible parties. These proposals concern legal reform rather than rights and obligations already comprehensively established under current law.