EVIDENCE IN DEEP LEARNING AND SYNTHETIC MEDIA CRIMES

Abstract

The rapid development of artificial intelligence, particularly deep learning technology, has significantly transformed digital communication and media production. One of the most controversial outcomes of this development is synthetic media, including deepfakes, AI-generated audio, manipulated videos, and fabricated visual evidence. While these technologies offer beneficial applications in entertainment, education, and digital innovation, they also create serious threats in criminal activities. Deep learning-generated content has increasingly been used in fraud, identity theft, defamation, cyber extortion, election manipulation, and digital impersonation. This phenomenon creates substantial challenges in evidentiary law because traditional methods of authentication are often insufficient to verify the originality and reliability of AI-generated materials.This paper examines the legal and evidentiary challenges in proving crimes involving deep learning and synthetic media. Using normative juridical research methods, this study analyzes statutory regulations, legal doctrines, forensic practices, and comparative legal frameworks from Indonesia, the United States, and the European Union. The study finds that the complexity of synthetic media requires stronger forensic methodologies, revised evidentiary standards, and adaptive procedural laws. It also highlights the urgent need for legal reform to establish clearer mechanisms for authentication, burden of proof allocation, and expert testimony in deepfake-related crimes. This research contributes to the discourse on criminal evidence law by proposing an integrated evidentiary framework that combines digital forensics, chain of custody principles, and AI-detection systems to preserve justice and procedural fairness.