Generative artificial intelligence (GAI) has taken the world by storm, but this excitement has come with difficult legal questions going to the very heart of the technology. In this article, we explore the concept of fair use, highlighting recent decisions from the U.S. Supreme Court and a key case that could shape the future of GAI.

Fair use and recent Supreme Court precedent

At the heart of GAI legal battles is the issue of whether training GAI technology by copying other works constitutes fair use, a legal doctrine that permits the unlicensed use of copyright-protected works in certain circumstances. When deciding whether a particular use is entitled to the fair use defense, courts balance these four factors:

The first factor—the purpose and character of the use—asks whether the use was “transformative”, but courts have long struggled with how to draw lines between what is and is not transformative. Some uses are more settled. For instance, a parody “transforms” the original work by satirizing it.

The Supreme Court found Google’s Android developers "transformed", not merely copied, Oracle’s code (Google LLC v. Oracle Am., Inc.)¹

Oracle accused Google of copying parts of Oracle’s Java application programming interface (API) to develop the Android operating system. Oracle demanded US$8.8 billion in damages for Google's sales and licensing of earlier Android versions that allegedly used the APIs without permission. While Google admitted to replicating 11,500 lines of Oracle's code, Google argued that this was done solely to enable developers to create applications for Android, not to replicate the functionality of Java itself.

The U.S. Supreme Court sided with Google. The Court found that Google “reimplemented”, not merely copied, Oracle’s API. By reimplementing the API, Google maintained consistency so programmers could use their acquired skills in a different context. The Supreme Court reasoned that such reimplementation of a functional interface furthered the development of computer programs and allowed the market to grow. Rather than just repurposing code “from larger computers to smaller computers”, the Court found that “Google’s Android platform was part of a distinct (and more advanced) market than Java software".² In the Court’s view, the fair use assessment should take into account those public benefits as compared to the dollar amounts likely lost by the copyright holder.

GAI developers will undoubtedly rely on this decision in arguing for fair use.

… but not every change in style makes the use fair (Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith)³

Building upon Google, the Supreme Court handed down its decision Warhol Foundation last year. This is a case the copyright holders will likely rely on in arguing that GAI is not fair use.

Photographer Lynn Goldsmith sued the Andy Warhol Foundation for the Visual Arts (AWF) in 2017 over the use of her 1981 photograph of Prince. The photograph, originally licensed to Vanity Fair for a US$400 fee, was used by Warhol to create a series of prints without Goldsmith's knowledge. Years later, AWF licensed one of these prints, Orange Prince, to Condé Nast for US$10,000, again without Goldsmith's permission.

The Supreme Court held that the Foundation’s licensing of Orange Prince was not a fair use of Goldsmith’s photograph. The transformativeness inquiry, the Court clarified, is one of degree, and merely adding new expression to a work does not necessarily make it transformative under fair use. So, licensing “Orange Prince” to a magazine was not considered transformative because both Warhol’s work and Goldsmith’s original photo were used for similar commercial purposes. Even though Warhol had added artistic expression, this did not change the fundamental purpose or commercial nature of the specific use at issue.

Copyright holders suing GAI developers will look to this decision to argue their works enjoy a licensing market: GAI developers’ use plays in the same licensing market and cannot be transformative for that reason.

GAI copyright litigation

Several lawsuits have been filed against companies whose GAI technologies have been trained on copyrighted materials.

Although these cases are far from deciding the central fair use question, there is one, much further along, that may decide the issue.

Thomson Reuters v. Ross Intelligence⁴ centers on the alleged unauthorized use of Thomson Reuters’s Westlaw platform. The platform offers a database of judicial opinions in a digital format, organized by legal discipline and annotated with short summaries of points of law.

Ross created a natural-language search engine that answers legal questions by returning relevant quotations from judicial opinions. Because the tool leverages GAI, Ross sought out legal material for training purposes. Some of this material came from the Westlaw platform, but the method by which Ross fed Westlaw’s information to its machine is in dispute.

Both parties presented arguments on fair use, but Judge Stephanos Bibas found that too many facts remained in dispute for him to conclude one way or another. For one, he explained the first factor, whether Ross’s use was transformative, could fall either way. On the one hand, Ross’s use was undoubtedly commercial. On the other hand, the transformativeness of Ross’s use weighed in favor of fair use. Citing Warhol Foundation, Judge Bibas recognized a “use’s transformativeness may outweigh its commercial character",⁵ but declined to indicate where this balance lied. Thus, this issue will be brought to a jury.

Another factor—the amount and substantiality of Ross’s alleged copying—will also go to the jury. The Google Court held that it “generally weigh[s] in favor of fair use where . . . the amount of copying was tethered to a valid, and transformative, purpose".⁶

 Thomson Reuters argued Ross copied excessively; Ross contended it needed a wide breadth of materials to train its AI.

Judge Bibas ended his fair use analysis with a question hitting at the core of the GAI copyright zeitgeist: “Is it in the public benefit to allow AI to be trained with copyrighted material?

How the jury perceives the public benefit of GAI may play a crucial role in the outcome of this case.

If the jury finds Ross’s technology is protected by the fair use defense, developers may be handed a formidable shield against copyright lawsuits threatening the future of GAI.

A finding that the technology is not fair use, however, could be a dagger in the hearts of GAI developers—and a sigh of relief for writers, artists and musicians alike.

‍

David Munkittrick is a partner and Perry Kumagai is an associate in the Litigation Department of Proskauer.