Abstract
The article analyzes the current model of civil liability for damage caused by artificial intelligence systems in the Republic of Kazakhstan and assesses the need for its improvement. The study examines why, despite the adoption of a special law on artificial intelligence, compensation for harm continues to rely on the traditional tort mechanisms of the Civil Code, and whether these mechanisms are capable of addressing algorithmic autonomy, complex technological chains, evidentiary asymmetry, and risk allocation. Using formal-legal, comparative-legal, systemic, and doctrinal methods, the research analyzes the Law of the Republic of Kazakhstan “On Artificial Intelligence”, the Civil Code, domestic legal doctrine, the European Union AI Act, the Product Liability Directive, and foreign legal practice. The study concludes that granting artificial intelligence independent legal personality does not simplify compensation mechanisms but instead obscures the identification of liable persons. A more effective approach is proposed whereby external liability toward the victim is concentrated on the entity that introduced and used the system, while liability among developers, integrators, and service providers is distributed through recourse claims. The article also argues that, for high-risk systems, failure to preserve logs, audit materials, and risk documentation should create adverse evidentiary consequences for the defendant.