The use of Artificial Intelligence in the field of criminal punishment – Chances, risks and constitutional limitations

This article reviews the prospects and constitutional constraints of Artificial Intelligence (AI) in criminal justice systems under EU, Regional, German, and Greek Law. After differentiating AI from conventional Legal Tech, the authors explore AI applications across the criminal process, especially in sentencing and risk assessment, and evaluate advisory to fully automated uses against legal standards. Potential benefits include less bias, clearer decisions, faster processing, and improved predictions. Risks involve system opacity, discrimination, automation bias, and reliance on general statistics over individual cases. The article highlights the EU AI Act's “high-risk” classification for criminal-justice AI, relevant transparency and oversight provisions, and national rights to human judgment. The authors argue that full automation in sentencing violates constitutional norms, but transparent, accountable AI can assist – without replacing – judges. Adoption requires a solid legal foundation, safeguards against bias, explainability, verifiability of input data, and real human control. The article urges interdisciplinary research, specialized training, and public debate for responsible AI integration in criminal justice.