The Snoopy Solution: How Fair Use and Licensing for Generative AI Can Coexist: Matthew Sag, Emory

This lecture argues that copyright safety measures in generative AI systems should be understood not merely as compliance requirements but as strategic opportunities for cooperation between AI developers and creative industries. Although the fair use doctrine protects nonexpressive uses of copyrighted works in AI training, AI developers find it hard to purge all vestiges of memorization from their models, especially where what counts as memorization is more abstract, such as with copyrightable characters (the famous "Snoopy Problem"). As such, both to stay on the right side of fair use and to avoid being held liable for specific impinging outputs, AI developers will need to implement various copyright safety measures. Rather than viewing these safeguards purely as restrictive impositions on the development of technology or the expressive freedom of users—which they are to some extent— we should also regard them as opportunities for licensing, a way out of the zero-sum conflict between technology and creative interests.

Matthew Sag is the Jonas Robitscher Professor of Law in Artificial Intelligence, Machine Learning and Data Science at Emory University Law School. Professor Sag is the founder of the Legal Scholars Roundtable on Artificial Intelligence, and he is a world-leading expert on the intersection of copyright law and text data mining, machine learning, and generative AI. He is also a sought after speaker on the broader implications of generative AI and its use as a legal technology. In July 2023, he testified to the U.S. Senate Judiciary Committee Subcommittee on Intellectual Property in relation to copyright and Generative AI and his research is published in leading journals such as Nature and Science, California Law Review, Emory Law Journal, Georgetown Law Journal, Northwestern Law Review, Notre Dame Law Review, and the Vanderbilt Law Review.