The framing of AI systems within the definition of product under defective products legislation An analysis of the current legislation with a focus on the Proposal for a Directive of the European Parliament and of the Council of 28 September 2022 (COM/2022/495)

Abstract

The current legislation on defective products, resulting from the transposition of Directive 85/374, aims to regulate damages caused by goods that lack safety, including artificial intelligence systems, until the recently approved Directive intended to replace it comes into force. This legislation is a powerful tool for ensuring adequate compensation for victims for the damages contemplated therein, which are increasingly common in a digitized society. However, those who seek to interpret and apply this legislation to artificial intelligence systems face a complex reality that is difficult to align with the general concepts and definitions set out therein. This paper specifically addresses the difficulties in classifying artificial intelligence systems (and their components) as products within the scope of the legislation. The study of this issue is rather complex for several reasons: it is not clear what should be understood by artificial intelligence; AI’s close dependence on data and certain services to perform some of its functions; and its open nature and ongoing modification through updates and improvements. In addition, AI systems can be marketed in different ways (either as a standalone product or as a component of another), which has consequences for the applicable regime. Lastly, AI systems themselves can be distributed or marketed as a service, requiring an analysis of whether the damages they cause are covered by the specific legislation on defective products.