The regulation of ai-based judges in judicial systems: international standards, guiding principles, and normative gaps: a scoping review

Abstract

The integration of artificial intelligence systems into the administration of justice has progressively transformed the judicial function, but its regulatory development has not kept pace. This scoping review aimed to map the academic and institutional production on international standards, guiding principles, and normative gaps in the regulation of so-called "AI-based judges." Following the PRISMA-ScR methodology, searches were conducted in ScienceDirect and Scopus, identifying 357 records, of which 15 original studies (2020-2025) were included after a three-phase selection process. The results show consensus around principles such as transparency, explainability, human oversight, fairness, and accountability, although their operationalization presents significant limitations. Critical normative gaps are identified: lack of liability criteria for AI-related harms, absence of technical standards for algorithm auditing, insufficient data protection rules adapted to the procedural context, and lack of judicial training policies. The empirical experiences analyzed—from e-justice platforms to risk assessment systems—reveal regulatory fragmentation across jurisdictions. It is concluded that the harmonization of standards, following the European AI Act model, is urgent to ensure an ethical and safe adoption of AI in justice.