Аннотация
The article examines the problems of legality and legitimacy of decisions of state authorities implemented through the use of artificial intelligence, with a particular focus on cases of algorithmic errors and the related issues of legal liability. It is noted that the introduction of AI technologies into public administration significantly transforms traditional mechanisms of administrative decision-making and necessitates a reconsideration of fundamental constitutional and legal principles, including the principles of legality, legal certainty, and the protection of human rights. The study substantiates the need to develop a theoretical and legal definition of an algorithmic error as a new legal phenomenon emerging at the intersection of automated technologies, data processing, and managerial decision-making, which is capable of exerting a substantial impact on the legal status of individuals. Algorithmic errors are considered not merely as technical failures, but as legally relevant events that may lead to violations of citizens’ rights and lawful interests. Based on a comparative legal analysis of foreign experience (the European Union, the United Kingdom, and the Netherlands), as well as the current regulatory framework of the Republic of Kazakhstan, the article proposes criteria for assessing the legality and legitimacy of decisions adopted with the participation of artificial intelligence. These criteria include the limits of permissible automation of public functions, requirements for human oversight, and procedural safeguards ensuring effective protection of rights. The article formulates a model of distributed legal liability among state authorities, public officials, and developers of AI systems, providing for both compensatory and recourse mechanisms for damage caused. Special attention is paid to substantiating directions for improving the legislation of the Republic of Kazakhstan aimed at introducing standards of transparency, explainability, and auditability of artificial intelligence, as well as strengthening procedural guarantees for the protection of citizens’ rights. The conclusions of the study contribute to the formation of a comprehensive model of legal regulation of artificial intelligence and to the strengthening of the principle of legality in digital public governance.