THE USE OF ARTIFICIAL INTELLIGENCE IN FORENSIC SCIENCE: CHALLENGES AND PROSPECTS
TL;DR
It is argued that AI should remain an auxiliary tool, while final forensic conclusions and procedural decisions must be made by competent human professionals in accordance with the law.
Abstract
This article examines the opportunities and challenges associated with the use of artificial intelligence (AI) in forensic science at the stages of collecting, examining, and evaluating evidentiary information. It considers the use of machine-learning algorithms, computer vision, and language models to process digital datasets, analyse photographs and video recordings, conduct biometric comparisons, examine documents, and detect synthetically generated content. AI can accelerate information retrieval and help identify relationships between data; however, technological efficiency does not guarantee the reliability of the result. The article analyses the risks of algorithmic bias, limited explainability, errors in training data, loss of digital context, breaches in the chain of custody, and excessive reliance on automated conclusions. In light of Ukrainian legislation and European approaches, it proposes a multi-level model for verifying neural-network outputs. The model includes documenting the tool, preserving original data, independent validation, describing methodological limitations, and ensuring that the parties to proceedings have a genuine opportunity to challenge the result. The article argues that AI should remain an auxiliary tool, while final forensic conclusions and procedural decisions must be made by competent human professionals in accordance with the law.
