CIVIL LIABILITY FOR DAMAGES CAUSED BY AUTONOMOUS DECISIONS OF ARTIFICIAL INTELLIGENCE SYSTEMS: A COMPARATIVE STUDY OF SAUDI, FRENCH, AND EUROPEAN UNION LAW

This study examines civil liability arising from damages caused by the autonomous decisions of artificial intelligence (AI) systems, employing a comparative analytical approach across three legal frameworks, namely, the Saudi legal system, the French Civil Code, and European Union regulatory initiatives. With the increasing reliance of AI systems on varying degrees of autonomy making decisions without direct human intervention in vital sectors such as healthcare, transportation, and finance traditional tort rules, historically based on human error, face unprecedented challenges. The central research questions are: to what extent can current civil liability frameworks accommodate damages resulting from algorithmic decision-making, and what legal reforms are needed to bridge the resulting liability gap? The study reveals that liability based on fault is insufficient when there is no clear human error behind an AI system's autonomous decision. The principles of joint and several liability presuppose a legal relationship between two persons, which cannot be established with a non-legal entity. The absolute liability of the custodian of things (garde des choses), as stipulated in Article (132) of the Saudi Civil Transactions Law and Article (1242) of the French Civil Code, provides the most viable traditional basis, but it faces difficulties in defining the custodian and the concept of effective control over autonomous systems. Product liability, significantly strengthened by the revised EU Directive on Product Liability (2024/2853), which now explicitly includes software, represents a promising complementary approach. From an Islamic legal perspective, the study identifies the fundamental principles of liability (daman/liability) that provide a solid jurisprudential foundation, particularly the principle that aggression (contagion) entails liability regardless of intent, and the standard model of liability for the actions of animals, which shares structural similarities with autonomous AI systems. This study proposes a comprehensive legislative framework for the Kingdom of Saudi Arabia, comprising six pillars: classifying AI systems based on risk, joint and several liability among stakeholders, the rebuttable presumption of causation, mandatory transparency, mandatory insurance for high-risk AI systems, and a supplementary compensation fund. These recommendations aim to strike a balance between protecting victims and fostering innovation, aligning with the goals of Saudi Vision 2030 and the Kingdom's designation of 2026 as the Year of Artificial Intelligence.