The rapid deployment of artificial intelligence (AI) within autonomous decision-making systems has generated profound challenges for traditional legal doctrines of liability, accountability, and responsibility attribution. This study examines the normative boundaries of AI liability through a doctrinal and normative legal research approach based on statutory regulations, international legal instruments, judicial decisions, regulatory frameworks, and contemporary legal scholarship. The analysis focuses on the evolving relationship between legal personhood, fault, foreseeability, risk allocation, and accountability within increasingly autonomous technological environments. The findings demonstrate that existing legal systems continue to reject the attribution of independent legal liability to AI systems despite their growing operational autonomy. Liability remains primarily attached to developers, deployers, operators, data controllers, and institutional actors whose decisions shape the design, governance, and deployment of AI technologies. Contemporary legal developments reveal a gradual shift from exclusively fault-based liability toward hybrid frameworks integrating risk management, preventive obligations, and accountability-based governance. The study concludes that the most coherent doctrinal model is a layered accountability framework in which responsibility is allocated according to governance capacity, control, risk creation, and regulatory obligations, thereby preserving legal certainty while accommodating technological transformation in autonomous decision-making ecosystems.