INTRODUCTION Contributed by: Dale Cendali, Joshua Simmons and Jeremy King, Kirkland & Ellis LLP
Kirkland & Ellis LLP 601 Lexington Avenue
New York NY 10022 USA
Tel: +1 212 336 4800 Fax: +1 212 446 4900 Email: jeremy.king@kirkland.com Web: www.kirkland.com
2025 was a year dominated by artificial intelligence, producing landmark court rulings in AI-based copy - right cases both in the United States and abroad. The past year also produced the Dewberry Group v Dew- berry Engineers Supreme Court decision in the United States, resetting the table for how plaintiffs determine which corporate entities to include as defendants. At the same time, trade mark offices across the globe are trending towards speedier, more efficient outcomes and cracking down on use requirements and over - broad registrations. AI In 2025, US courts and those abroad issued highly anticipated rulings at the intersection of copyright and artificial intelligence. In the United States, district court decisions in Thomson Reuters v ROSS Intelli- gence , Kadrey v Meta and Bartz v Anthropic provided guidance to litigants as to the paramount importance of a plaintiff showing the effect on the market for works used in training under the United States’ fair use rubric. Emphasising the importance of the effect on the mar - ket factor, the Thomson Reuters court held that the defendant’s copying of Westlaw headnotes to create a competing legal research product was not fair use. This case is currently on interlocutory appeal in the United States’ Third Circuit Court of Appeals, where it stands to be the first US appellate decision applying copyright law to the training of an artificial intelligence. Both Kadrey and Bartz were class actions brought by authors asserting that the defendants infringed plaintiffs’ copyrights by ingesting the plaintiffs’ works to train the defendants’ respective AI models. Key to
both decisions, however, was there was no evidence that either defendant’s AI product was capable of cre - ating infringing works, placing the focus of the fair use inquiry on the defendants’ use of the authors’ works to train the AI model. Although, as in Thomson Reuters , the Kadrey court confirmed that the most important factor in a fair use analysis is the effect of the infringing work on the market, it ruled against the plaintiffs on the facts presented. The Kadrey court noted that the plaintiffs could have shown a market effect by the compet - ing works generated by defendant’s Large Language Model (LLM), “even if the model can’t regurgitate [plaintiffs’] own works or generate substantially simi - lar ones”, if the LLM “can generate works that are similar enough (in subject matter or genre) that they will compete with the originals and thereby indirectly substitute for them”. The court, however, held that these particular plaintiffs had failed to do so. The court ultimately concluded that the plaintiffs had only argued “harm from the loss of fees paid to license a work for a transformative purpose” which it described as “not cognizable”, holding that the defendant’s use was a fair use. The opinion’s discussion of market effects by competing works, however, offers a trail of breadcrumbs to would-be plaintiffs on how to present a more compelling damages case on similar facts. In Bartz v Anthropic , the court held that even if a market for AI training data for LLMs were to develop, “that use is not one the Copyright Act entitles Authors to exploit”. The court, however, drew a distinction between the facts in its case and Thomson Reuters , agreeing with the Thomson Reuters court that “using a proprietary system for finding court opinions in
6 CHAMBERS.COM
Powered by FlippingBook