Legal Challenges of Generative AI in Cybercrime Investigation

ABSTRACT

The proliferation of generative artificial intelligence (“generative AI”) tools has transformed both the practice of cybercrime and the evidentiary foundations upon which its investigation and prosecution depend. Large language models and synthetic-media generators now enable low-skill offenders to produce convincing phishing communications, malicious code, and audio-visual deepfakes at a scale and quality previously reserved for sophisticated actors. This paper examines the legal challenges that generative AI poses to cybercrime investigation, focusing on three interlocking problems: the authentication and admissibility of AI-generated or AI-altered digital evidence; the adequacy of existing criminal-procedure and evidence law to address synthetic-media manipulation; and the fragmented state of cross-border cooperation for the collection and sharing of electronic evidence. Employing a doctrinal and comparative legal methodology, the paper analyses developments in the United States (proposed Federal Rule of Evidence 707 and amendments to Rule 901), the European Union (the Artificial Intelligence Act’s transparency and law-enforcement provisions), India (the 2026 Information Technology Rules amendments governing “synthetically generated information”), and the newly opened United Nations Convention against Cybercrime. The paper finds that while jurisdictions are converging on the need for provenance-based labelling, expedited takedown regimes, and burden-shifting authentication rules, substantial gaps persist in forensic capacity, judicial training, and mutual legal assistance mechanisms, creating an accountability lag between technological capability and legal response. The paper concludes with recommendations for harmonized evidentiary standards, investment in forensic-AI detection infrastructure, and rights-respecting international cooperation frameworks.