Artificial Intelligence in Courts, Arbitration and Issues of “Hallucinations”

Abstract

The article examines specific aspects of artificial (machine) intelligence in dispute resolution in state and arbitration courts. Drawing on legal research methods and data from artificial intelligence science, the authors propose a significant obstacle to its application in the legal sphere is the so-called “hallucination problem,” i.e., the generation of false data by machine intelligence programs that appear entirely reliable. The objectives of the article are: a) to define and determine the properties of artificial intelligence and its generative variety; b) to identify the specific features of artificial intelligence application in judicial proceedings and arbitration; c) to examine foreign judicial practice in which false data generated by artificial intelligence programs has been identified; and d) to formulate the problem of artificial intelligence “hallucinations” and to seek ways to overcome it. The authors conclude that arbitration is a dispute resolution method more receptive to technological innovations, including artificial intelligence. The potential for implementing machine intelligence in state courts is less clear. However, litigants are using AI tools in their trial preparations. International judicial practice clearly demonstrates the propensity of AI to “hallucinations”: judges uncover fabricated data during trials, leading to liability for litigants who use AI programs without due diligence. Two solutions have been proposed to overcome the problem of “hallucinations.” The first is technological and draws on the work of Russian scientists, who propose the development of more sophisticated generative AI programs free of “hallucinations.” The second solution involves thoroughly reviewing legal texts and procedural documents for false data in cases where there is reason to believe a document was drafted using machine intelligence.