Abstract
Respect for the rights of persons taking part in interrogation remains one of the weakest links between codified criminal-procedure guarantees and law-enforcement practice, because the written protocol, rather than an objective record, is still the primary evidence of how a statement was obtained; this gap undermines access to justice, human rights, and strong institutions. The study examines whether, and under what legal conditions, algorithmic analysis of audiovisual and textual data from interrogations can serve as a verification tool for procedural compliance in the Republic of Kazakhstan. It employs formal-legal analysis of national legislation, including the 2025 Law on Artificial Intelligence, and international standards; a comparative-legal matrix of recording and oversight regimes in Kazakhstan, the United States, and England and Wales/the European Union; and conceptual modelling of a compliance-assessment algorithm implemented in a functional edge-computing prototype. The results show that Kazakhstan codifies all eight compared guarantees but backs only three of them with a mandatory objective record, against six in England and Wales and four in United States federal practice; twelve procedural guarantees were mapped onto fourteen machine-verifiable indicators, of which nine are deterministic and five probabilistic; the prototype captures four synchronized Full HD (1920×1080, ≥25 fps) streams and performs on-device speech recognition on an 11 TFLOPS, 32 GB module. The findings suggest that most procedural guarantees can be verified by rule-based analysis of recording metadata, transcript and video without recourse to deception or emotion inference. In contrast, probabilistic indicators can only support, never replace, human oversight of peace, justice and strong institutions. JEL Classification Codes: K14, K41, K42, O33, O38.