Artificial intelligence in Ukrainian traditional categories of criminal law

Abstract

At the request of the International Association of Criminal Law (AIDP-IAPL, Association International de Droit Pénal – a non-governmental organization on criminal law, Paris, France) within the framework of the XXI International Congress of Criminal Law "Artificial Intelligence and Criminal Justice", a subgroup of the Ukrainian national group of AIDP-IAPL in two scientists prepared detailed answers to questions about ▪ definition and legal qualification of Artificial Intelligence system (legal definition of AI system in Ukrainian law, Ukrainian AI-based systems for predictive policing, legal definition of Machine Learning in Ukrainian law, legal personhood or legal capacity to the AI systems), ▪ existing criminal offences and criminalization (illegal act committed by, through or against an AI system; new offences related to designing, programming, developing, producing, functioning or making use of AI systems; crimes of mere conduct, commission and omission offences, consummate offence, crimes with intent, etc.; who can be considered the possible perpetrator and / or victim of the new AI offences e.g. producers / programmers / system engineers / developers / designers etc.; specific mental element of individual criminal liability, possibility for legal persons be held liable for AI crimes committed by any person acting individually or having a leading position within the legal person; criminal responsibility of the perpetrator or of the legal person in order to avoid the risk of over- criminalization if the AI systems are produced, used or put on the market for legal purposes, e.g. for scientific or research reason; whether reports or legal literature suggest the introduction of new criminal offences linked to AI systems; positive obligations for persons and/or legal person designing, developing, producing, testing, selling or distributing AI systems etc.), ▪ applicability of Traditional Criminal Law Categories (possibility for AI system considered as a “computer system” as defined by Article 1, lett. of Cybercrime Convention and / or Article 2, lett. a) of Directive EU/2013/40; specific problems with respect to the principle of legality; admissibility of analogy in order to criminalize illegal acts related to AI systems; joint-perpetrator or participant in the commission of the crime; forms of secondary liability applicable to AI-related crimes; state of mind (e.g. dolus) on the part of the human agent who designed / programmed / developed / produced / circulated / marketed / used the AI system; exact and concrete modus operandi of the AI system in committing the offence etc.), ▪ adaptation of Traditional Criminal Law Categories and academic debate (principle of culpability nullum crimen sine culpa and mens rea; compliance with the principle of culpability when the output causing the harm generated by the intelligent machine is neither wanted nor predictable by the human agent; compliance with the principle of culpability when an AI system is intentionally used by a human agent as a tool but the AI system carried out an offence different from the one wanted by the human agent; criminal participation and attempted crimes; liability of legal persons; necessary adjustments of the legal principles on criminal liability of legal persons when they are involved in AI-related crimes; necessary adjustments of policies and preventive measures within private organizations in order to guarantee a correct and regular use of AI systems etc.), ▪ alternatives to criminalization and non-criminal sources.