Technology of Deepfake: Transformation of Evidence Standards in Court Proceedings

Abstract

The authors examine changes in the institution of judicial evidence caused by the widespread use of generative artificial intelligence and deepfakes technologies, the use of that in judicial practice is considered a complex challenge for the modern legal system. Fake photos, videos, and audio materials can be used to falsify evidence, and verifying the authenticity of such evidence requires the involvement of digital forensics experts. The experts analyze file metadata, look for compression artifacts, inconsistencies in lighting, shadows, or unnatural movements of the eyes and facial muscles. The authors analyze why the classical elements (criteria) of the standard of proof — relativity, admissibility, and authenticity — are losing their value in today’s conditions, when any audio or video information can be completely generated by a neural network and is visually indistinguishable from the original. Three key shortcomings of the traditional standard of proof are discussed: relevance does not protect against synthetic forgery, admissibility does not guarantee the authenticity of evidence, and the property of evidence’s reliability is unattainable due to the lack of approved methods for detecting deepfakes. The need to shift from “proof of facts” to “proof of provenance” of evidence is substantiated. A verifiable chain of custody is proposed as a new element of the standard of proof, including mandatory metadata recording, cryptographic hashing of files, and documentation of all stages of storage and transfer of a digital object. In the absence of such a chain, it is proposed to introduce a presumption of evidence generation. Based on a comparison of international legal experience (USA, European Union, China) and Russian judicial practice, particular attention is paid to the inadmissibility of using entirely AI-generated materials as independent evidence in legal proceedings, as well as the admissibility of the analytical use of neural networks in forensic examinations. The article may be of scholar and practical purpose for judges, investigators, lawyers, forensic experts, and researchers in the field of digital technology integration in law enforcement and evidence law.