12th Annual Peter A. Jaszi Distinguished Lecture on Intellectual Property

Although we are still a long way from the science fiction version of artificial general intelligence that thinks, feels, and refuses to “open the pod bay doors”, recent advances in machine learning and artificial intelligence (“AI”) pose profound challenges to copyright law and the fair use doctrine. The challenges posed by generative AI show why, now more than ever, we need a theory of fair use that reflects fundamental copyright principles and eschews considerations of broader social welfare that federal judges are poorly qualified to assess. I argue that expressive substitution is the key to understanding and applying fair use and that this leads to a wide affordance for non-expressive uses of copyrighted works. In previous work I have suggested that, in the context of non-expressive use by copy-reliant technology, the absence of direct expressive substitution is all that is required for fair use. Recent developments in generative AI suggest that non-expressive use may not be the be all and end all of fair use, however. In this Article I argue that in addition to considering direct expressive substitution, courts assessing whether training machine learning programs on copyright works is fair use may also consider whether the challenged use undermines the economic incentives that copyright is designed to create.