Rischio e Intelligenza Artificiale. Un’analisi concettuale tra razionalità e incertezza

This article questions the nature of the threats related to the use of AI systems and asks whether it is correct to consider these issues under the category of ‘risk’. Our thesis is that the concept of risk associated with AI is understood in a purely modern sense, i.e. as calculable and predictable, but the kind of risks that these systems pose do not lend themselves to such an interpretation. The argument will therefore be structured in three parts. Firstly we will propose a reconstruction of the history of the concept of risk and the evolution of its mean- ing. From there, we will focus on the specificity of the risks associated with the use of AI and their relationship with human decision–making and knowledge. Finally, we will offer some conclusions by referring to the risk–based approach adopted by the European regulation on the subject (AI Act), setting out what – in our opinion – are some critical aspects of this perspective.