IA: modelli tradizionali di responsabilità e nuove prospettive

The essay explores the issue of civil liability for damages caused by artificial intelligence in light of Roman legal tradition and its subsequent doctrinal developments. Starting from the fre-quent comparison between the “intelligent machine” and the servus of Roman law, the author highlights its inadequacy: the status of the slave, a person subject to dominica potestas, cannot serve as a meaningful parallel for attributing liability to artificial entities. Instead, the analysis iden-tifies in Roman law the origins of strict liability, constructed through an absolute presumption of culpa in vigilando or in eligendo imposed on the person responsible for supervising others. This model – rooted in the inseparable link between iniuria and culpa – differs fundamentally from modern vicarious liability, which, following the natural law theories of Thomasius and Pothier, in-troduced a relative presumption of fault and thus a subjective form of liability. The essay further examines contemporary European legislative initiatives, particularly the AI Liability Directive, which tends to objectify liability while maintaining the traditional fault-based structure, mainly by easing the burden of proof on the injured party. Finally, the author discusses the proposal to at-tribute “electronic personality” to intelligent machines, underscoring its largely symbolic rather than substantive character. The study concludes that modern trends toward the objectification of liabil-ity, though differently framed, revive the original Roman rationale of protecting the injured party by assigning responsibility to the subject best positioned to bear the cost of harm.