Acceso usuarios
Barcelona, España
This paper aims to explore a novel aspect of Directive (EU) 2024/2853 (PLD): the manufacturer’s liability for continuous product learning. This is a new criterion for evaluating product defectiveness that inte-grates artificial intelligence. The PLD pays limited attention to this matter, so the study aims to identify the conditions under which a product’s continuous learning might be deemed defective. Firstly, it is argued that this basis for assessing whether a product is defective should be understood to apply to generative AI systems, rather than to AI systems that do not change their behaviour after deployment. Secondly, the ten-year limitation period for holding the manufacturer liable, starting from the time the product was placed on the market or put into service, seems unsuitable for products that undergo continuous learning, as their features may change over time. Thirdly, the crucial factor enabling the manufacturer’s liability is that the product remains within the manufacturer’s control, allowing modifications to correct defects caused by continuous learning.
