EXTRA The timing was mere coincidence, and the context different. But the official organs or the copyright industries were quick to pounce on what is at least a colorable connection between the Supreme Court’s 7-2 ruling in Andy Warhol Foundation v. Lynn Goldsmith and the current bubbling controversy around the use of copyrighted works to train generative AI models.
“We applaud the Supreme Court’s considered and thoughtful decision that claims of ‘transformative use’ cannot undermine the basic rights given to all creators under the Copyright Act,” Recording Industry Association of America CEO Mitch Glazier said in a statement. “We hope those who have relied on distorted – and now discredited – claims of ‘transformative use,’ such as those who use copyrighted works to train artificial intelligence systems without authorization, will revisit their practices in light of this important ruling.”
“Today’s decision reinforces the essential role that copyright plays in society,” added Maria A. Pallante, President and CEO of the Association of American Publishers. “The Court’s decision is consistent with the points we raised in our 2022 amicus brief. Importantly, it reaffirms the fact that transformative use under the first fair use factor requires a robust analysis about the use at issue and cannot be interpreted so broadly as to swallow the derivative work right.”
The Warhol case stemmed from a series of silkscreen images of Prince produced by the artist but based on a photograph taken by Lynn Goldsmith. Vanity Fair magazine licensed the photo for use as an artist’s reference and then commissioned Warhol to create an portrait of the singer to illustrate a story following Prince’s death. In a later issue commemorating Prince’s life and work, the magazine licensed a second image from Warhol’s series (left) without obtaining reuse rights from Goldsmith.
After the Goldsmith notified Vanity Fair that she had not authorized the use of her photograph for a second image Warhol’s estate filed suit seeking a declaratory ruling that the entire silkscreen series qualified as a fair use of Goldsmith’s portrait. The district court ruled in favor of Warhol but was reversed on appeal.
The Supreme Court has now affirmed that reversal.
“Goldsmith’s original works, like those of other photographers, are entitled to copyright protection, even against famous artists.” Justice Sonya Sotomayor wrote in her majority opinion. “Such protection includes the right to prepare derivative works that transform the original. The use of a copyrighted work may nevertheless be fair if, among other things, the use has a purpose and character that is sufficiently distinct from the original. In this case, however, Goldsmith’s photograph of Prince, and AWF’s copying use of the photograph in an image licensed to a special edition magazine devoted to Prince, share substantially the same commercial purpose.”
The Court heard oral arguments in the case back in October, and many court observers had expected a ruling long before last week. But judging by the pointed, almost personal tone of the exchanges between the majority and dissent in footnotes to their respective opinions, the case was clearly contentious within the Court itself.
The rough consensus among outside experts is that the Court’s actual holding is fairly narrow in a legal sense. In rejecting the Warhol Foundation’s claim of fair use the majority focused on its commercial purpose in licensing the second silkscreen print to the magazine, rather than on any creative or transformative use the artist may have made of Goldsmith’s photograph. Nor did it spend much time addressing any of the other factors that go into a typical fair use analysis.
Indeed, as Justice Neil Gorsuch wrote in his concurring opinion, “The Court today does not even decide whether the Foundation’s image of Prince infringes on Ms. Goldsmith’s copyright. To uphold a claim of infringement under the Copyright Act, a court must find the defendant copied elements of the plaintiff ’s work that are themselves original… The plaintiff must usually show not only a similarity but a ‘substantial’ similarity between the allegedly infringing work and the original elements of his own copyrighted work… In this case, we address none of these questions or other elements of the infringement standard designed to ensure room for later artists to build on the work of their predecessors.”
Narrow, or not, the ruling lands just as the U.S. Copyright Office is taking the initial steps toward a formal inquiry into, among other things, the use of copyrighted works to train generative artificial intelligence models. As part of that process, the office has been hosting a series of listening sessions among stakeholders in which the question of transformative fair use has featured prominently, and at times contentiously, in the discussions (see here and here).
Rights owners accuse AI developers of wholesale illegal copying of their works in the training of generative AI models. Developers counter that their models are not making perceptible copies of their training data and their use of copyrighted material is transformative and should therefore be regarded as a fair use.
In support of their position, developers frequently cite the Google Books case (Authors Guild et. al. v. Google 2d Circuit 2015). In that case, the Second Circuit Court of Appeals ruled that Google’s copying of the plaintiffs texts to create the searchable Google Books database was transformative, in that it resulted in a new product that advanced the constitutional purpose of copyright, and therefore qualified as non-infringing fair use.
In the first of the USCO’s listening sessions, Emory University professor of law and artificial intelligence was quick to invoke the ruling in defense of AI developers’ use of copyrighted material in training their models. “The test for infringement is copying, in fact, and substantial similarity, and that remains the same, no matter how a work is created,” he said. “The copying required to collect the training data for these large language models is a classic form of non-expressive use that was upheld as fair use in Google Books.”
That ruling, which the Supreme Court let stand as precedent, has been a thorn in the side of copyright owners since the Second Circuit handed it down. Nonetheless, Copyright Alliance CEO Keith Kupferschmid did his best to try to distinguish AI training from the Google case.
“The Google Books case could not be more different from what we have going on here,” Kupferschmid insisted. “Google did not copy books to make new books. That’s what AI does… In Google Books, Google used the works for informational purposes. They use it for the information in the works, not the expressive content of the works. That is exactly what AI is doing. They’re using the expressive content to produce new works.”
Technical issues around the operation of generative AI models aside, the Supreme Court’s ruling in Warhol, now gives copyright owners a precedent that, rhetorically at least, can stand against the Google Books case. Whatever else generative AI models like Midjourney and ChatGPT may be doing, their developers are unquestionably making commercial use of them. And insofar as their use can result in new works that potentially compete with the material they were trained on, that commercial purpose could weigh against a finding of fair use in light of Warhol.
Certainly RIAA’s Glazier reads it that way.
“Transformative” appears nowhere in the text of §107 of the Copyright Act defining fair use. It’s a concept read into the law by judges and therefore in the eyes of the beholder. But it’s now likely be the most hotly contested legal ground in the battle over the use of copyrighted works to train AI models. And both sides are now equally well armed.