An Approach to the Instantiation of the EU AI Act: A Level of Done Derivation and a Case Study from the Automotive Domain

Fabian Hüger, Alexander Poth, Andreas Wittmann, Roland Walgenbach

Abstract: Based on the EU AI Act draft from November 2022 a team of data scientists, quality managers and legal experts set out to instantiate the AI Act for their project domain. To focus on the product and service relevant parts of the extensive EU Act, the Level of Done (LoD) layer approach was applied. Based on this LoD layer for the AI Act an evaluation was initiated with ongoing Machine Learning (ML) projects. This case study describes the method and approach on how the instantiation was done and provides a first insight into the LoD-application from an engineering perspective.