The Evidentiary Value of AI-Generated Data: A Framework for Reliability and Admissibility

Abstract

The increasing use of artificial intelligence (AI) in criminal investigations has introduced complex evidentiary challenges. AI-generated data—such as facial recognition results, predictive risk assessments, and automated surveillance outputs—often lack transparency, auditability, and human accountability. This paper examines whether such data should be admissible as evidence in criminal trials, and under what conditions. It argues that traditional evidentiary rules are insufficient to address the unique risks posed by opaque and autonomous sys-tems. Drawing on comparative insights from multiple jurisdictions, the article proposes a normative framework based on three principles: verifiability, pro-cedural accountability, and proportionality. The goal is to ensure that efficiency gains from AI do not override fundamental rights such as the presumption of innocence, the right to a fair trial, and the ability to challenge adverse evidence. The paper concludes by offering concrete recommendations for legal reform, including disclosure requirements, independent technical review, and stricter judicial reasoning when admitting AI-generated evidence.