1* Marek VOCHOZKA
Abstract: Autonomous artificial intelligence (AI) is reshaping legal doctrine by blurring the lines of agency and accountability. This study asks two intertwined questions: first, how current liability regimes can be recalibrated to address harms caused by self-learning AI; and second, whether recognizing AI as a limited legal subject offers a viable path to meaningful accountability. Drawing on doctrinal analysis, comparative jurisprudence, and philosophical inquiry, the paper examines the EU AI Act (2024/1689), the EU AI Liability Directive, the GDPR, and parallel developments in U.S. and Canadian law. A systematic survey of many peer-reviewed sources and recent case law reveals persistent regulatory gaps: liability is still funneled through human actors, strict-liability models remain scarce, and key terms such as “autonomous” and “high-risk” lack harmonized definitions. The research highlights new frameworks, such as partial legal capacity (Teilrechtsfähigkeit) and the Robotics/AI Legal Entity (RAiLE), that suggest a hybrid model of AI personhood. However, it finds no jurisdiction willing to grant full legal status to AI. Liability chains often break when AI systems go beyond their original programming. This leaves victims under-protected and developers unsure of their risks. This study ends with incremental, adaptive reforms, stranded in distributed agency, clearer constitutional language, and anticipatory governance, which are critical to defend justice and public confidence as intelligent systems turn more profoundly embedded in social and economic life.
Keywords: Artificial Intelligence, Legal Responsibility, Algorithmic Accountability, AI Personhood, Liability Frameworks
JEL codes: K13; K20; K24.
Received: 10 Jul 2025; accepted: 02 Mar 2026; published: 30 June 2026.
How to Cite This Article: Vochozka, M. (2026). AI and Legal Accountability: Rethinking Responsibility in Algorithmic Decision-Making. Littera Scripta, 19(1), 85-102. https://doi.org/10.36708/ LS.2026.I01.006