Abstract

At last count, thirty lawsuits are challenging the legality of using in-copyright works to train generative AI models. While all but one of the lawsuits are still in early stages, fair use defenses are likely to be the main focus of judicial analyses of their merits. This talk will explain what judges have decided so far, that is, what issues are now out of those cases and what issues remain. Although generative AI developers point to some precedents that support their fair use defenses, this talk will explain why judges may find them not as helpful as the developers hope. At this point, some cases seem stronger than others. Not only are the plaintiffs seeking very large damage awards, but also some want courts to order models trained on the plaintiffs' data to be destroyed, and courts have power to issue such orders. The broader impacts of these lawsuits for academic research and development are at present invisible to the courts.