인공지능 증거의 심리방법

Examination of AI-Generated Evidence

The era of artificial intelligence has arrived, sparking discussions on how AI can be utilized in litigation processes. However, there has been relatively little discussion about how to evaluate and examine evidence generated by AI when it is submitted in court. This paper begins by examining how different jurisdictions address the admissibility and weight of evidence, as a foundation for studying the examination of AI-generated evidence. It argues that rejecting AI evidence at the admissibility stage is unreasonable. Furthermore, after considering the unique characteristics of AI-generated evidence, this paper explores the possibility of subjecting such evidence to cross-examination. Following a discussion of the Daubert standard, which applies to general scientific evidence, the paper focuses on and analyzes key issues—such as error rates, data size, and the “black box” nature of AI—that may arise when applying the standard to AI evidence.