Criminal liability and artificial intelligence

Responsabilidade penal e inteligência artificial

Abstract

This article aims to relate legal personality and responsibility related to intelligent artificial entities. Its hypothesis is that not only general AI or superintelligence (still non-existent) can theoretically commit criminal acts: even common AI systems currently available can be trained by criminals to commit acts that violate fundamental rights. And culpable situations (due to negligence, recklessness or malpractice of programmers, designers and directors of AI development companies, for example) must be considered in the event of crimes involving AI. At least currently, therefore, alternatives more consistent with current Criminal Law should be considered for criminal situations involving AI, such as new criminal types in the current law against criminally responsible humans, or further emphasize civil punishment for damages caused ​​by autonomous artificial entities. Specifically, its objectives are: i) to present the possibilities of criminal use of AI, as well as the legal impossibility of attributing criminal liability to artificial entities endowed with this technology at the current moment of technological evolution; ii) to explore new opposing theoretical and legal possibilities — in other words, what should be done for, one day, when there is a general AI, endowed with as much awareness as a human being and, therefore, perhaps guilty. Methodology: hypothetical-deductive procedure method, qualitative approach and bibliographic review technique.