Criminal law is limited by the legality principle, which can lead to some difficulty when apprehending new types of offenses. Lately, one of the struggles is the development of Artificial Intelligence and its possible interaction with criminal law. One of the main difficulties regarding crim-inal law categories is the lack of clear definitions but, more importantly, the categorization of AI as the object or the subject of a criminal offence. In this article, we will look at the definition problem to conclude on the need for a sectoral approach, before looking more into the categorization of AI. As an object, a.k.a. as a complex tool to facilitate or to commit an offense, the traditional categories seem sufficient to apprehend the concept. For the latter issue, traditional categories seem to be lacking in the possibility of an AI committing an offense or a legal act as a subject, a.k.a. by itself. However, if civil law or commercial law is prompt on accepting AI as a legal person, criminal law still ponders the possibility.