Legal liability for artificial intelligence actions: challenges and regulatory prospects

The article examines the issue of determining legal liability for damage caused by artificial intelligence (AI) systems, given their growing autonomy and ability to make decisions without direct human intervention. It analyses current approaches to regulating such legal relations in the EU, the US and Ukraine, as well as the trend towards the formation of a global model that strikes a balance between technological innovation and the protection of human rights. Considerable attention is paid to the issue of the distribution of liability between developers, manufacturers, data providers, owners, operators and users, as well as to discussions on the possibility of granting autonomous systems certain elements of legal personality. The article outlines the peculiarities of AI application in high-risk areas – autonomous transport, medicine, finance, security, and public administration – where technical failures or algorithmic bias can cause significant damage or human rights violations. A comparative analysis of liability models is conducted: strict (objective), joint, proportional and insurance, which are used in leading jurisdictions. Based on the EU’s experience, the key provisions of the AI Act are examined, which introduces a risk-based control system and establishes requirements for transparency, data management, cybersecurity, and oversight of high-risk systems. The features of the American model, which is based on the doctrines of product liability and negligence and allows judicial authorities to respond quickly to technological challenges, are highlighted. It is noted that in Ukraine, the absence of specific legal regulation of AI creates significant uncertainty, primarily in matters of compensation for damage and establishing causality. The necessity of updating the Civil Code of Ukraine, introducing algorithm certification, risk insurance, data auditing, and increasing the transparency of developers’ activities has been substantiated. A hybrid approach is proposed that takes into account the level of autonomy of systems and provides adequate protection for victims without restricting technological development. The results of the study can form the basis for the development of a comprehensive regulatory model that will ensure the safe and ethical use of AI in Ukraine.