Copyright protection during the training stage of generative AI: Industry-oriented U.S. law, rights-oriented EU law, and fair remuneration rights for generative AI training under the UN's international governance regime for AI

Abstract

Generative AI relies on simulating and learning from complex data distributions to automatically generate new, meaningful content from large datasets. Training generative AI models carries the risk of copyright infringement. How can we balance the development of generative AI technology with copyright protection during the training stage of these models? In contrast to existing legal scholarship, this article conducts a comparative study of industry-oriented U.S. copyright law and rights-oriented EU copyright law. It draws insights from the concepts of the tragedy of the commons and the tragedy of the anti-commons, proposing the creation of fair remuneration rights for Generative AI training under the UN's international governance regime for AI. This article offers a typological analysis of potential operational models for fair remuneration rights in Generative AI training, drawing analogies to existing remuneration rights in international copyright treaties. The goal is to provide an open framework for further discussion within the international academic community.