The EU Legal Choice Architecture (CA) for General-Purpose AI Models (GPAI Models)

Abstract

Abstract This chapter explores the legislative choice architecture established by the European legislature with respect to the development, providing and deployment of generative AI. Focusing on the legal requirements that apply to real-world AI applications, the chapter presents and analyzes fundamental notions introduced by the AI Act, in particular general-purpose AI model, AI system and general-purpose AI system. Based on these notions, the chapter focuses on two normative frameworks that are crucial for generative AI: data protection law and the provisions introduced in AI Act concerning general-purpose AI models presenting systemic risk. Section “Requirements for the Architectural Choices to Be Made by Developers, Data Controllers and Processors, Providers and Deployers” engages in a reflection on the implications of the requirements established by the AI Act and data protection law for Legal Protection by Design. By discussing the obligations to carry out a Fundamental Rights Impact Assessment and a Data Protection Impact Assessment, the chapter illustrates how EU law establishes a choice architecture that encompasses the whole value chain of generative AI.

Ähnliche Arbeiten