In the context of digital transformation and exponential data growth, traditional methods of prosecutorial supervision are losing their effectiveness. The integration of Artificial Intelligence (AI) technologies opens up new opportunities, yet breeds complex procedural and ethical conflicts, including the risks of technocratic arbitrariness and algorithmic discrimination. To develop a scientifically grounded doctrine for the application of AI in the supervisory activities of the Prosecutor General's Office of the Republic of Kazakhstan, considering the latest legislative restrictions, and to define strict boundaries for algorithmic intervention. The study employs a comprehensive approach, including dialectical, formal-legal (analysis of the Law of the RK "On Artificial Intelligence" and relevant concepts), and comparative-legal methods to contrast the Kazakhstani, Asian, and Anglo-Saxon regulatory models. It is established that a strictly regulated "human-centric" AI model is emerging in Kazakhstan. Three levels of limits for algorithm application are substantiated: epistemological (AI acts exclusively as an assistant), procedural (the inadmissibility of replacing a prosecutor's inner conviction with a machine-generated conclusion), and ethical (legislative prohibition of manipulation and biometric discrimination). The study proves the necessity of adapting criminal procedural legislation to the norms of digital law, implementing a sovereign IT infrastructure, and developing prosecutors' competencies in supervising algorithmic legality. Meeting these conditions will ensure a balance between technological progress and the strict protection of citizens' constitutional rights.