Jane Tullis
A Delaware jury may shortly be called upon to adjudicate a precedential case regarding the intersection of artificial intelligence, copyright law, and fair use. Amidst the overall hype surrounding AI and the debate surrounding fair use since its inception, this jury will be faced with a monumental task.
In May of 2020, Reuters brought a copyright infringement suit against ROSS Intelligence with regard to ROSS’s alleged scraping of copyrighted material on Westlaw. Reuters accused ROSS of using Westlaw’s “West Headnotes” and “West Key Number System” to train a new machine learning legal research tool.[1] Upon the completion of this new search engine, ROSS would then be able to compete with Reuters’s Westlaw without having expended the same resources. Reuters specifically alleged that they would be, “in the unfair position of having to compete with a product that they unknowingly helped created.”[2] In response, ROSS attempted to assert a fair use defense. In their argument, they stated that their models only scrapped ideas and facts, which are not copyrightable, and their purpose was to create new code, as opposed to the creation of a competing product.[3] In September of 2023, the judge presiding over the case denied both parties’ motions for summary judgment and instead called for a jury trial, including on the issue of fair use.[4]
Attempting to adjudicate an emerging technology has always been challenging, but this case is complicated further by the need to simultaneously assess a fair use defense. Courtrooms have struggled to accommodate the increasing presence of scientific advances in litigation, especially because of the dynamic nature of so many of these developments.[5] How is a jury member supposed to determine the facts of a case, weigh them against each other, and then reach a conclusion without being able to fully understand the technology at play? In a scientifically invalid survey of my law school peers, their explanations of artificial intelligence understandably ranged from “AI synthesizes data to create outputs” to “evil scalper” and “machine learning!” Despite being bombarded with mentions of AI—and cautionary tales to avoid writing case briefs using ChatGPT—there is still not a uniform understanding of how large language models function and rely on copyrighted material to do so.
While judges have similarly struggled with the incorporation of science into the courtroom, the addition of a fair use defense only heightens the potential for confusion amongst jury members.[6] Regarding the role of the jury in a fair use trial, copyright expert and law professor David Nimmer highlighted the complexities associated with fair use and the difficulty associated with creating jury instructions that fully encapsulate fair use without bogging down the jury with unnecessary details.[7] Though Nimmer ultimately calls out the “elitist disdain for jury determinations of fair use,” he recognized that even judges, who interact much more frequently with copyright law than juries, still struggle with the tenets of intellectual property law and the constitutional principles that animate it.[8] Through no fault of their own, jury members are tasked with reaching a conclusion on the basis of a doctrine that has inspired decades of academic debate and confusion (and law review blog posts).
While juries absolutely play a critical role in our legal system, this case emphasizes the current challenge set before them. There are several similar cases being litigated across the country, though they lag behind this particular case by years.[9] This jury will simultaneously have to grapple with potentially unfamiliar technology and wade through the potential merits of a hotly-debated copyright defense. Their challenges will only be compounded by the precedential nature of the case. The instructions given to the jury and the ultimate conclusion of the jury can impact how attorneys, creators, and software programmers approach the use of copyrighted material in large language models for years to come. What, like it’s hard?
[1] Complaint at 11, Thomson Reuters Enter. Centre GmbH v. ROSS Intel. Inc., No. 1:20-cv-00613 (D. Del. May 6, 2020).
[2] Id. at 1.
[3] Reuters’ Generative AI Copyright Case Must Be Heard by Jury, Per DE Court, The Fashion Law (Sept. 28, 2023), https://www.thefashionlaw.com/issues-in-reuters-generative-ai-copyright-case-must-be-heard-by-jury/ [https://perma.cc/7MLL-X6QX] [/web/20231005034206/https://www.thefashionlaw.com/issues-in-reuters-generative-ai-copyright-case-must-be-heard-by-jury/].
[4] Thomson Reuters Enter. Centre GmbH v. ROSS Intel. Inc., No. 1:20-cv-00613, 2023 WL 6210901 (D. Del. Sept. 25, 2023).
[5] Honorable Jed S. Rakoff, Science and the Law: Uncomfortable Bedfellows, 38 Seton Hall L. Rev. 1379, 1392–93 (2008).
[6] Id. at 1379–80.
[7] David Nimmer, Juries and the Development of Fair Use Standards, 31 Harv. J. Law & Tech 563, 588 (2018).
[8] Id. at 588–89.
[9] The Fashion Law, supra note 3.