Artificial intelligence and the defect of synthetic judicial reasoning in drafting judicial decisions

Abstract

The use of generative artificial intelligence (AI) in judicial chambers raises the question of which tasks judges may assign to these systems without compromising the constitutional duty to state reasons. This article aims to delimit the tasks that may be assigned to AI in drafting judicial decisions and to characterize the defect of synthetic or artificial judicial reasoning that arises when such assistance replaces the judge's decisional operations. Through a documentary review with a doctrinal and comparative legal approach, the study examined Peruvian legislation, international standards, 5 foreign judicial decisions, and a corpus of 51 institutional documents submitted by 22 Peruvian Superior Courts of Justice, gathered through Official Communication No. 000001-2026-P-CSJAR-PJ. Results show that none of the 22 Superior Courts analyzed has specific internal regulations on AI use in decision drafting, that tool adoption remains incipient and fragmented, and that the comparative case law reviewed has annulled decisions grounded on AI-hallucinated content. An attribution-and-incidence test is proposed to determine the invalidity of judicial reasoning, together with a Traceability and Judicial Validation Protocol governing task allocation, verification, and accountability. The article concludes that AI may draft even a complete first draft without vitiating judicial reasoning, provided the judge has already determined the proven facts, assessed the evidence, and defined the outcome, and subsequently subjects the text to substantive review.