Is It Fair Use to Use Pirated Materials for AI Training?

This panel will discuss the current controversy over whether the use of “pirated” datasets in training AI models overcome claims of fair use. The panelists will discuss the differing treatment of this issue in Bartz v. Anthropic, Kadrey v. Meta, and In re: OpenAI, and whether training AI models on pirated works furthers copyright’s goals of promoting public knowledge while incentivizing the creation of new works. The panel will explore how various U.S. Supreme Court decisions have assessed the role of weighed the stated intent of the user against its actual use in the fair use analysis, and how those decisions might impact the fair use analysis in future decisions. The panel will also discuss whether authorized access to copyrighted works for AI training should play a threshold role in fair use analysis, including how the European Union addresses that question in its text and data mining exceptions under the 2019 CDSM Directive.