Problems of Recognizing the Results of Generative Artificial Intelligence as Other Documentary Evidence in Criminal Proceedings

The present study examines the issue of procedural qualification of the outputs produced by generative algorithmic systems (ChatGPT, Claude, Gemini) from the perspective of evidentiary law. Central attention is devoted to analyzing the extent to which the products of such technologies comply with the criteria for other documents within the framework of Russian criminal proceedings. The research identifies the essential characteristics of materials generated by these models, demonstrating their fundamental incompatibility with classical forms of documentary evidence and revealing the systemic contradictions that arise when attempting to qualify such materials under Article 84 of the Criminal Procedure Code of the Russian Federation. It is proved that generative models create materials devoid of the main feature of evidence – connection with past circumstances, representing probabilistic constructions synthesized based on statistical patterns of training data. Five key problems of procedural qualification of such materials are examined: establishing relevance to the circumstances of the case, ensuring admissibility of evidence, verifying the reliability of contained information, absence of possibility to interrogate the author, and distinguishing between evidential and reference information. The paper proposes a conceptual solution to the problem of distinguishing between evidential and non-evidential information in the context of using generative artificial intelligence materials in criminal proceedings, and substantiates the necessity of legislative consolidation of the prohibition on recognizing the results of automated systems as other documentary evidence.