Algorithmic justice in the making: AI, sentencing, and fairness in Malaysia’s criminal courts

Abstract

Artificial intelligence (AI) is increasingly used in criminal justice systems to improve efficiency and consistency, but its use in sentencing raises important questions about fairness, transparency, and judicial discretion. This article examines Malaysia’s introduction of AI-assisted sentencing through the Artificial Intelligence in Court Sentencing (AiCOS) system and asks how algorithmic assistance reshapes the exercise and visibility of sentencing discretion. It adopts a qualitative, desk-based, and interdisciplinary approach, drawing on criminology, legal and sociolegal scholarship, judicial and institutional materials, and international AI governance frameworks. The Malaysian experience is examined through four dimensions: transparency and explainability, procedural contestability, human discretion, and accountability, with comparative reference to the United States’s experience with COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) and the European Union’s regulatory framework. The article argues that AiCOS is best understood not as an autonomous ‘AI judge’, but as an automated evolution of Malaysia’s existing Table of Sentencing . While its guidance function remains broadly similar, algorithmic mediation changes the visibility and contestability of the basis upon which punishment is determined. Automation therefore does not eliminate discretion but may relocate it into less visible decisions concerning data, system design, deployment, and reliance on algorithmic recommendations. The article concludes that retaining a human judge as the final decision-maker is insufficient, by itself, to ensure fairness. Effective governance requires transparency, disclosure, meaningful opportunities to challenge algorithmic inputs and recommendations, reasoned judicial decision-making, and continuing independent oversight.