In the new chapter of technological development, the emergence of various artificial intelligence models has profoundly affected numerous fields, including arbitration. Traditional or even online methods of commercial dispute resolution, in light of the expansion of electronic transactions and the massive volume of documents, have rendered the speed, accuracy, and efficiency of evidence evaluation inadequate. The purpose of this research is to examine the criteria governing evidence verification and to answer the following question: What is the standard for the use of artificial intelligence in evidence authenticity assessment in arbitration, and how can it effectively realize that standard? Current studies indicate that the prevailing standards in the evaluation of evidence in arbitration are the balance of probabilities and clear and convincing evidence. Using a descriptive-analytical approach and through analysis of the subject, this research ultimately concludes that, given the dominance of party autonomy and the flexible nature of evidentiary procedures in arbitration, the criteria and standards of proof vary according to the will of the parties. Artificial intelligence, by employing fuzzy logic, induction, and objective criteria, is capable of implementing conventional standards of evidence verification—such as the balance of probabilities—in a practical and effective manner. The absence of explicit and specific criteria and standards in existing laws, the probabilistic and data-driven nature of AI inferences, and the resulting ethical and legal challenges constitute major obstacles in the use of artificial intelligence for evidence verification in arbitration. Therefore, technical measures to develop and improve artificial intelligence algorithms, as well as the enactment of laws and regulations to fill the current regulatory vacuum, are essential.

Key words: standard of proof, induction, analogy, fuzzy logic, validation of evidence.