EU Copyright Law, Artificial Intelligence and ‘Transformative Use’ of Works: The Case of 3D Reconstruction

Abstract

Artificial intelligence (AI) has virtually upended many concerns of EU copyright law. In parallel, three-dimensional (3D) visual content is predicted to be a key contributing factor to the development of new forms of immersive content, much of which is powered by new AI-driven tools. AI-based ‘3D reconstruction’ techniques are unique because, among others, they may autonomously generate an individual piece of 3D visual content, based on specific 2D visual content used as an input. This raises particular copyright questions regarding the copyright-relevant uses of such techniques.

This contribution analyses these forms of use, and the extent to which the acts performed on underlying 2D visual content as training material are copyright-relevant under EU copyright law. In particular, the analysis specifically tackles how those uses interfere with the exclusive right of reproduction, its exceptions and limitations at the input stage, and whether a relation of derivation between 2D and 3D content may be established at the output stage. We highlight avenues through which the training and implementation of 3D reconstruction techniques may in fact not be copyright-relevant and analyse the possibilities through which the implementation of 3D reconstruction techniques may qualify as a form of ‘transformative use’ of 2D visual content, including whether this can be subject to exceptions and limitations.