Az algoritmikus büntetéskiszabás lehetőségének jogi megfelelőségi vizsgálatáról

The study examines the legal compliance of algorithmic sentencing, focusing on its possibilities and limitations. The research is based on the statistical and substantive analysis of 541 final judgments in human trafficking cases. The findings reveal inconsistencies in sentencing practices, where the impact of mitigating and aggravating circumstances is often unpredictable. This justifies the exploration of algorithmic models, particularly in simpler cases where factors can be clearly classified and scaled. The paper outlines both symbolic and non-symbolic AI models and their potential role in supporting sentencing decisions. Special attention is given to the legal compliance framework, combining EU law (AI Act, LED Directive, GDPR) with Hungarian legislation. The author emphasises that algorithms may only serve as decision support tools, while the final judgment must remain the judge’s responsibility. The analysis also addresses human rights and constitutional requirements, including judicial independence, fair trial rights and the prohibition of discrimination. The main objective is to establish a compliance matrix applicable to the assessment of future algorithmic sentencing models. Ultimately, the study contributes to a critical understanding of the intersection between legal science and artificial intelligence.

Keywords: algorithmic sentencing, artificial intelligence, legal compliance, data protection (LED, GDPR), fair trial, human trafficking