Artificial intelligence, risk regulation, and civil liability: a legal analysis of the EU AI Act

This essay examines the risk management framework introduced by the Artificial Intelligence Act and focuses on the classification of AI systems by risk level and the corresponding preventive obligations, highlighting the absence of a comprehensive civil liability regime in the event of harm. Particular attention is given to the notion of residual risk, i.e. the risk that persists despite compliance with the regulatory safeguards. In this context, the essay explores the potential applicability of Article 2050 of the Italian Civil Code, which governs liability for dangerous activities, arguing that mere compliance with the AI Act does not suffice to exempt the user of high-risk AI systems from liability. The essay concludes by calling for complementary legislative intervention to provide a clearer framework for civil liability, capable of ensuring effective protection for injured parties while also fostering technological innovation.

Il contributo analizza il rapporto tra regolazione del rischio e responsabilità civile nell’ambito del Regolamento europeo sull’intelligenza artificiale (AI Act). Dopo aver ricostruito l’impostazione risk-based del Regolamento e la classificazione dei sistemi di IA in base al livello di rischio — inaccettabile, elevato, limitato e minimo — l’analisi si concentra sugli obblighi preventivi previsti per i sistemi ad alto rischio, tra cui la gestione continua dei rischi, il controllo dei dati, la documentazione tecnica, la tracciabilità, la valutazione d’impatto sui diritti fondamentali e il monitoraggio successivo alla commercializzazione.