CRIMINAL LIABILITY FOR HARM CAUSED BY ARTIFICIAL INTELLIGENCE SYSTEMS: CROATIAN LEGAL REFORM AND THE EU LAW CONTEXT

Abstract

The rapid deployment of artificial intelligence (AI) systems across the European Union (EU) raises profound legal challenges, particularly regarding liability for damage caused by autonomous or semi-autonomous systems. While AI promises significant economic and societal benefits, its increasing autonomy disrupts traditional legal concepts of fault, causation, and responsibility. This tension is clearly illustrated by recent national debates, such as in Croatia, where proposals have emerged to introduce criminal liability for harm caused by AI systems, resulting in new criminal offence introduced in 2025, despite unresolved questions concerning the identification of the responsible legal subject. At the EU level, policymakers have sought to address these challenges through a comprehensive regulatory framework, most notably the Artificial Intelligence Act (AI Act) and the proposed reforms of civil liability rules for defective products and emerging digital technologies. These instruments aim to balance innovation with legal certainty, risk management, and the protection of fundamental rights. However, they predominantly focus on ex ante risk regulation and civil liability mechanisms, leaving unresolved issues regarding criminal liability and the attribution of responsibility in cases involving complex AI-driven decision-making processes. This paper examines the evolving EU approach to liability for AI-caused damage and assesses its implications for legal certainty within member states. The research focuses on three core questions: (1) how EU law conceptualizes responsibility and accountability for AI-related harm; (2) how national legal systems, using Croatia as a case study, attempt to transpose or supplement EU norms through criminal and civil law instruments; and (3) whether existing regulatory models adequately address the practical and normative challenges posed by autonomous AI systems. Methodologically, the paper employs doctrinal legal analysis of EU legislation and policy documents, comparative examination of national legal approaches, and critical review of relevant academic literature. Where available, emerging judicial reasoning and analogies from existing case law on technological risk and product liability are also considered. The paper argues that current regulatory developments reveal a fragmented approach to AI liability that risks undermining legal certainty and uniform application of EU law. It concludes by proposing normative guidelines for a more coherent EU-wide framework that clarifies the attribution of responsibility, enhances harmonization across member states, and strengthens the EU’s role as a global standard-setter in AI governance.