Second Circuit Rejects Internet Archive’s Controlled Digital Lending Scheme
For a second time, IA’s fair use defense is soundly rejected by a federal court
On September 4, a panel of judges in the U.S. Court of Appeals for the Second Circuit issued a decision in Hachette v. Internet Archive, affirming a district court’s categorical rejection of the Internet Archive’s (IA’s) fair use defense that it attempted to use to excuse its contrived Control Digital Lending (CDL) scheme. The decision is not just a big win for book publishers but also for the copyright community—and, as the Court points out, the public—as it provides a clear explanation of what the provisions of the Copyright Act permit and how they relate to the incentives that drive creators to produce new works. There are even aspects of the decision that provide some insight into how the Second Circuit might apply the four-factor fair use analysis to the generative AI cases that may eventually be before it.
Background
A detailed summary of the procedural background and other materials can be found here, but the short version is that back in 2020, four book publishers sued IA for copyright infringement for scanning tens of thousands of their copyrighted literary works and distributing those digital copies to the public for free. IA defended its activity by invoking a fabricated theory called controlled digital lending (CDL), which it claimed made available no more digital copies of a book than the number of print copies of the book in its possession (what it described as a one-to-one “owned to loaned ratio”). In March 2023, the District Court for the Southern District of New York granted summary judgment in favor of the publishers and, in doing so, denounced Internet Archive’s CDL theory and soundly rejected its fair use defense, finding all four fair use factors favored the publishers. Then, in September of 2023, IA appealed the decision to the U.S. Court of Appeals for the Second Circuit.
IA’s Use is Unequivocally Not Transformative
The Court begins its fair use analysis by examining the purpose and character of IA’s use, focusing first of the question of whether the use is transformative. The decision cites heavily to the Supreme Court’s recent Warhol v. Goldsmith opinion, recognizing that while the right of copyright owners to prepare derivative works is subject to fair use, an “overbroad concept of transformative use . . . that includes any further purpose, or any different character, would [improperly] narrow the copyright owner’s exclusive right to create derivative works.”
The Court rejected IA’s argument that its use of the publishers’ works is transformative because it makes book lending “more convenient and efficient” and enables uses not possible with physical borrowing, concluding that “neither of these justifications render IA’s use of the works transformative because the underlying purpose of making the Works available in a derivative format is still the same.” As for IA’s improved efficiencies point, the Court said that IA doesn’t provide any efficiencies beyond what the publishers already provide through their own eBooks. It’s worth noting that if this analysis was applied the ultimate purpose of a use in the generative AI context, it’s hard to see how AI companies could argue that the ultimate purpose of the ingestion of copyrighted material is any different than the purpose those works serve—i.e. for an end user or consumer to enjoy a work.
The Court goes on to explain that the digital versions of books IA produces serve the exact same purpose as the publishers’ works, which is “making authors’ works available to read.” And thus, the court found that “IA’s Free Digital Library is meant to―and does―substitute for the original Works” and that while not every instance of non-transformative use is clear cut, “this one is.” The Court provides this succinct conclusion to its transformative analysis:
In sum, because IA’s Free Digital Library primarily supplants the original Works without adding meaningfully new or different features that avoid unduly impinging on Publishers’ rights to prepare derivative works, its use of the Works is not transformative.
Second Factor Analysis Weighs in Favor of Publishers and Hurts AI Companies’ Fair Use Arguments
Analyzing the nature of the copyrighted works at issue under the second fair use factor, the Court rejects IA’s argument that this factor weighs neutrally because it copied and distributed both fiction and nonfiction books. Calling that argument an “oversimplification” of the concept that the second factor favors fair use when the works copied are factual (rather than fictional), the decision explains that even nonfiction books “represent the authors’ original expressions of those facts and ideas.”
This last point is an important clarification that may play a pivotal role in future generative AI fair use decisions. Some AI companies are claiming that the only thing they are copying when ingesting copyrighted works for training is unprotectable data or facts about those works. However, as the Second Circuit recognizes, when a work that mixes original expressive elements with unprotectable material is copied in full, it’s impossible to argue that all that’s being copied is the unprotected elements.
In the generative AI context, simply because a work is converted into a format that can more easily be ingested by an AI system does not mean that the work suddenly ceases to include copyrightable expression or loses copyright protection. What some AI developers consider to be unprotectable data or facts are actually protected copyrightable expression, and the Second Circuit seems to understand that.
IA’s Wholesale Copying is Not Justified
Considering the amount and substantiality of the use under the third fair use factor, the Court first notes that it is sometimes necessary to copy the entirety of a work to achieve a legitimate, transformative secondary purpose. The decision then compares IA’s use to the copying of books that occurred in the HathiTrust and Google Books cases, explaining that in those cases the wholesale copying of works was necessary to achieve the purpose of creating a transformative search function and snippet view.
IA argued that the third factor should weigh in its favor because copying the entire work was necessary to implement CDL, but the Court wasn’t buying it. The decision explains that scanning entire works for the purpose of creating substitutes for the publishers’ print and eBooks is not a transformative use and therefore copying entire works is not necessary under the third factor. Moreover, unlike HathiTrust or Google Books, IA distributes identical copies of the publishers’ works to the public. The Court takes little time to conclude that factor three weighs strongly in favor of the publishers.
IA’s Use Harmed the Potential Market for and Value of the Publishers’ Works
Moving to the fourth factor, the Court makes clear that it not only considers the harm caused by the alleged infringers specific actions, but also the market harm that would occur if such actions became widespread. Then, quoting Warhol, it explains that the question is not “whether the second work would damage the market for the first…but whether it usurps the market for the first by offering a competing substitute.” This idea of substitution or supplantation, the Court explains, is what closely relates the first and fourth factors. Importantly, the Court reaffirms that it is the burden of the party asserting the fair use defense to prove that the secondary use does not compete in the relevant market.
IA’s main argument is that its use of the publishers’ works cannot serve as a substitute for the originals because it offers a distinct service that constitutes a transformative use of the works. The Court dismisses this argument quickly, noting that IA’s digital copies serve the same purpose as the originals and are therefore likely to compete in the market as a substitute for the original—something that was not true of the copying in HathiTrust or Google Books. The decision also points out that the IA itself advertised its digital books as free alternative copies to the publishers’ print and eBooks.
After explaining why IA’s expert witness testimony of lack of market harm was flawed and “ill supported,” the Court agrees with the publishers’ assertions that they (1) have suffered market harm due to lost eBook licensing fees and (2) will suffer market harm in the future if IA’s practices were to become widespread. It explains that “were we to approve IA’s use of the Works, there would be little reason for consumers or libraries to pay Publishers for content they could access for free on IA’s website.” The Court, responding to IA’s complaint that the publishers have not produced empirical data, emphasized that the burden of proof is borne by the party asserting the defense and not the plaintiff-publishers.
Any Public Benefits are Outweighed by Market Harm
Addressing IA’s argument that any market harm should be balanced against the “significant public benefits” provided by its conduct, the Court concludes that both the publishers and the public will benefit if IA’s use is denied. Citing to the Supreme Court’s decision in Harper & Row v. Nation Enterprises, the Court notes that any infringer can claim it’s benefitting the public simply by increasing access to a copyrighted work. It then offers this explanation of what it sees as “short-term” benefits:
Within the framework of the Copyright Act, IA’s argument regarding the public interest is shortsighted. True, libraries and consumers may reap some short-term benefits from access to free digital books, but what are the long-term consequences? If authors and creators knew that their original works could be copied and disseminated for free, there would be little motivation to produce new works. And a dearth of creative activity would undoubtedly negatively impact the public. It is this reality that the Copyright Act seeks to avoid.
The Court also makes an important point about the exclusive rights of a copyright owner that allow them to set the terms of engagement for a limited time. It explains that “[t]his monopolistic power is a feature, not a bug, of the Copyright Act.” The word monopoly is sometimes used in a derogatory manner by those who take issue with the scope of copyright protection, but the Court here confirms that it’s a foundational aspect of copyright law that best promotes the progress of the arts.
Conclusion
While it recognizes that eBook licensing fees may be burdensome to some and limit access to creative works, the Court makes explicitly clear that authors have a right to be compensated in connection with the copying and distribution of their original creations. In a final rebuke to the Internet Archive, the Court says:
IA asks this Court to bless the large scale copying and distribution of copyrighted books without permission from or payment to the Publishers or authors. Such a holding would allow for widescale copying that deprives creators of compensation and diminishes the incentive to produce new works. This may be what IA and its amici prefer, but it is not an approach that the Copyright Act permits.
The Second Circuit’s understanding of what the Copyright Act permits and how that impacts incentives to create new works is enlightening. And the decision is likely to have a far-reaching influence, as the Internet Archive has already attempted (and been sued for) applying a similar infringement model to other types of copyrighted works. When one considers how the Court analyzed each of the fair use factors, it’s also a decision that may have an impact on various pending AI cases—especially those where AI companies argue that they are only copying unprotectable elements. That will certainly be something to track in the future.
If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!