Daniel Rodriguez Maffioli (Duke University School of Law) has posted “Copyright in Generative AI training: Balancing Fair Use through Standardization and Transparency” on SSRN. Here is the abstract:
The rapid evolution of Generative Artificial Intelligence (GAI) has brought about transformative changes across industries, often raising challenging questions surrounding data rights, especially within the context of copyrighted content. This paper delves into the nuances of the relationship between GAI and the fair use doctrine, highlighting the complexities that emerge when copyrighted data serves as the backbone for the development of large-scale AI models. By combining Benjamin Sobel’s training data taxonomy with the distinct stages of the Generative AI cycle, a hybrid framework is presented, offering a more granulated perspective on the applicability of fair use in GAI contexts. Recognizing the inherent limitations of the current legal paradigms, the paper introduces actionable proposals, emphasizing the need for enhanced transparency, data provenance measures, and the implementation of Standardized Data Licensing Agreements (SDLAs). Such measures aim to bridge the gap between AI developers and copyright holders, facilitating smoother negotiations and fostering trust. While the core discussion revolves around the interplay of GAI and fair use, the paper acknowledges broader policy challenges in the AI domain, urging for continuous exploration. Overall, this work underscores the necessity of adaptive, collaborative, and transparent strategies in harmonizing the objectives of innovation with the imperatives of intellectual property rights in the GAI landscape.