Civil Liability for Artificial Intelligence-Induced Harm in Indonesia: Reconstructing Fault, Causation, and Evidentiary Responsibility through a Control-Risk-Evidence Model

Abstract

The deployment of artificial intelligence in financial services, electronic commerce, health care, transportation, education, employment, and data-driven decision-making expands the possibility of harm that can no longer be explained solely through direct human conduct. This article examines the adequacy of Indonesia's civil-liability framework for losses caused or mediated by AI systems and develops a reconstruction model compatible with tort, breach of contract, consumer protection, electronic transactions, and personal-data protection. Using normative legal research with statutory, conceptual, and limited comparative approaches, the article finds that Articles 1365, 1366, and 1367 of the Indonesian Civil Code remain doctrinally relevant but encounter serious attribution and evidentiary gaps where AI systems are opaque, adaptive, distributed across multiple actors, and technically documented by defendants rather than victims. The article proposes a Control-Risk-Evidence Model that allocates responsibility according to practical control, economic benefit, capacity to prevent harm, and access to evidence. The model combines documentation duties, rebuttable presumptions of causation, stricter liability for high-risk deployment, and contribution-based recourse among responsible actors. It therefore preserves the corrective and preventive functions of civil liability without treating AI either as a new legal person or as a universal excuse for human and corporate actors.