Artificial Intelligence and Legal Accountability: Challenges for Modern Legal System

Abstract

The rapid diffusion of artificial intelligence (AI) into judicial administration, notarial practice, law enforcement, and financial services has exposed a structural gap between the autonomous, opaque, and self-learning character of algorithmic systems and the anthropocentric assumptions embedded in classical doctrines of legal responsibility. This article examines the extent to which fault-based liability, strict liability, and vicarious liability doctrines remain adequate to address harm generated by AI systems, and evaluates comparative regulatory responses in the European Union, the United States, Singapore, and Indonesia. Employing normative legal research through statutory, comparative, and conceptual approaches, this study finds that existing Indonesian instruments principally the Electronic Information and Transactions Law and the Personal Data Protection Law regulate AI only tangentially and leave algorithmic accountability, auditability, and causation largely unaddressed. By contrast, the European Union's risk-based Artificial Intelligence Act imposes graduated obligations of transparency and human oversight according to risk classification. The article argues for a hybrid accountability model combining risk-tiered liability rules, mandatory algorithmic auditing, statutory human-in-the-loop requirements for high-risk systems, and a dedicated compensation mechanism, as a normative foundation for future Indonesian AI legislation