Current criminal procedure legislation lacks legal definitions of artificial intelligence systems (AIS), criteria for their admissibility, and procedural guarantees for participants in legal proceedings. This, coupled with the active implementation of these technologies, creates risks of violating citizens' rights. The relevance and significance of this study stems from the need to develop a conceptual model that would legitimize the use of AI without undermining the fundamental principles of criminal procedure. The purpose of this article is to propose a legal framework, the implementation of which would ensure the safe integration of AI systems into criminal proceedings. The methodological basis of this study is based on formal legal, systemic-structural, and comparative legal approaches, ensuring the consistency of the proposed solutions and their compliance with the principles of criminal procedure. The scientific novelty lies in the development of the author's concept, which proposes a transition from the linear, hidden, and unregulated use of AI systems to a model for their differentiated application. Unlike existing approaches that view AI as a single tool, this paper substantiates the distinction between AI systems and "AI assistants" and "autonomous AI agents," with a differentiation of their legal status. For the first time, a "forced consensus" mechanism is proposed, transforming the interaction between a law enforcement agent and an algorithm into a formalized dialogue, generating a new procedural document—a dialogue protocol. The possibility of voluntarily transferring a special legal procedure to an autonomous AI agent while preserving the right to informed refusal is also substantiated. It is concluded that the introduction of AI into criminal proceedings is possible without undermining its fundamental principles only through the establishment of a system of procedural guarantees and additional legal mechanisms for interaction between humans and artificial intelligence systems.
Aram Edvardovich Arzumanyan (Thu,) studied this question.