Abstract: In response to the challenges and prominent limitations faced by the traditional theory of criminal procedural justice in the artificial intelligence era, the academic community has put forward the theory of "technical procedural justice" centered on "technical regulation". While this theory addresses the core dimension of rectifying judicial biases, it has not yet developed into a comprehensive and appropriate theoretical system. The typical characteristics of criminal procedure in the AI era are reflected in two aspects: on the one hand, the datatization of case evidence generation and intelligent assistance in litigation proof present distinct features of technicalization and objectification; on the other hand, technicalization has not completely eliminated the influence of human subjective factors, and litigation proof still retains certain probabilistic characteristics. Therefore, to achieve the reasonable determination of case facts, the theory of criminal procedural justice should integrate "technical cognition" and "judgment based on common sense and rationality" to formulate corresponding procedural norms. Within this new theory of criminal procedural justice, several measures are of crucial significance: improving the system of expert assessors, introducing common-sense review of conclusions generated by intelligent auxiliary decision-making systems, strengthening judges' subjective initiative in out-of-court investigations, and implementing the majority decision-making mechanism under the collegial panel system.