Criminal Infringement on Artificial Intelligence Systems: A Criminal Law Description

Abstract

Most specialists agree that artificial intelligence (AI) is the technology that defines our present and future. Intelligent technologies are becoming increasingly common in critical social spheres, such as security, energy, medicine, law enforcement and judiciary, as well as transportation. An AI system is an executable program code or coefficients of a model that, given certain input data, make the system produce a certain result. As any other computer information, an AI system may become an object of criminal infringements. The authors study infringements on AI systems that incur criminal liability under the current Russian legislation. They also single out such a type of infringements on AI systems as adversarial attacks. These attacks are cases when a user, knowing about the specifics of developing and creating an AI system, intentionally feeds it data that lead to the incorrect functioning of this system. Such methods of interfering with intelligent systems may not contain formal attributes of offences that are criminally punishable under current Russian legislation. The authors of the article prove that such actions may have a high degree of public danger, which is sufficient for their criminalization. They conclude that current Russian criminal legislation does not encompass all methods of publicly dangerous infringements on AI systems. A high degree of public danger associated with adversarial attacks means that they should be recognized as criminal infringements on AI systems committed through influencing them without the use of malware or unauthorized access.