Publication Details
Abstract
Objective: This article examines the criminal-law foundations of liability for crimes committed with the participation of AI, with particular attention to the status of AI, mens rea, algorithmic negligence, the ‘black box’ problem, causation, and the distribution of responsibility among developers, operators, users, and other human actors. Method: Using doctrinal, comparative-legal, formal-legal, systemic, and analytical methods, the article argues that the attribution of criminal responsibility to AI itself is incompatible with the anthropocentric structure of contemporary criminal law. Results: At the same time, treating AI exclusively as an ordinary instrument may fail to capture the specific risks created by autonomous and self-learning systems. Under this approach, AI may be treated as a technologically autonomous participant or ‘quasi-subject’ in the factual mechanism of an offence, but criminal responsibility remains attributable exclusively to natural persons or other legally recognized human-controlled entities on the basis of guilt, foreseeability, control, and causation. Novelty: The article therefore develops a human-centred model of distributed responsibility. The article further proposes mandatory technical audit trails, differentiated liability for AI developers and operators, and legislative recognition of technologically mediated and hybrid forms of criminal conduct.