Tag: fair use

Chamber of Progress: Old Rationales for a Brave New World

old rationales

The Chamber of Progress launched an initiative called the “Generate and Create” campaign to “defend fair use” and “promote AI creativity.” I don’t know whether they bought this campaign used from the basement of Fight for The Future or Electronic Frontier Foundation, but the following statement is worn-out rhetoric that sounds even weaker defending AI as a mode of production than it was defending online platforms as a mode of distribution:

To combat the growing legal and policy copyright threats against generative artificial intelligence, Chamber of Progress announced a new campaign, Generate & Create, highlighting the creative benefits of generative artificial intelligence and supporting established fair use protections for AI training and output.

The pro-creator message is a remix of a remix of Lessig’s “remix culture” argument against online copyright enforcement—a narrative which begat the “we’re all creators” argument against copyright rights. Instead of YouTube enables creators to break free of “gatekeepers,” now it’s AI enabling the same emancipation, though as discussed in this post, it’s hard to fathom who the “gatekeepers” are this time.

Meanwhile, the promise to “defend fair use” is code for “we’re funded by Big Tech to tilt at windmills while we lose legal arguments.” One does not “defend fair use” the way one defends a right to read banned books or a right to reproductive healthcare in the same states that like to ban books. There is no legislative agenda to abolish or amend Section 107 of the Copyright Act. Fair use is a balancing test courts apply in certain copyright infringement cases, and on the headline question as to whether machine learning (ML) with copyrighted works is exempted by fair use, there is nothing “established” about that answer despite CoP’s implication to the contrary.

Although fair use cannot be applied generally (i.e., it is a case-by-case consideration), it is true that all the copyright infringement claims against the various AI developers arise from the same general conduct and, therefore, invite similar or identical fair use defenses. Cutting to the final chapter, if Open AI loses to New York Times and Udio loses to UMG et al. in the Second Circuit, those outcomes are likely to be controlling on the fair use question of ML. Even if any of these cases goes to the Supreme Court, the likelihood of a reversal of an opinion out of the Second Circuit—so prolific on fair use case law—is a bet I wouldn’t make.

Nevertheless, the argument will be presented, and it goes something like this:  Gen AI breeds new creative works, in part by breaking down “barriers” for would-be creators, and because this productivity is consistent with the purpose of copyright, ML serves a transformative purpose and is, therefore, fair use. Notwithstanding the fact that a defendant can win on the transformative question and still lose on fair use overall, I suspect the AI developers may find their very expensive machines described by the courts’ precedent language as “slightly transformative.”

But AI is revolutionary! you might say. How can it be only “slightly transformative?” Answer:  for the same reason the Internet Archive’s Open Library is “slightly transformative”—because its purpose was a substitute for licensed ebooks. What is different about GAI, of course, is that it is generally a substitute as a mode of production more than as a mode of distribution, and to complicate matters, some professional creators are using AI tools and deriving benefits from those uses. So, if that sounds like the answer is “it depends,” welcome to the fact-intensive nature of the fair use defense, which cannot be broadly “defended” in the sense the CoP proclaims.

CoP et al. will promote the argument that because GAI fosters the production of more “creative works,” this predicted increase in output fulfills the purpose of copyright law. But the reason I put “creative works” in quotes is that for every 100 sound recordings to come out of an AI product like Udio, somewhere between an unknown and zero percent of those sounds will be “creative works” as a matter of law. Copyright only protects human authorship of creative expression, and that doctrine will not—and should not—change. Meanwhile, the question as to what the human creator must do in collaboration with GAI for the human to claim copyright in the resulting work is an evolving doctrine—one that is several years, and several lawsuits, away from becoming guidance.

With a product like Udio or Suno, where the business model depends on consumers generating music with a few simple prompts, it is fair to assume that the vast majority of the music produced will not be “creative expression” as a matter of law. And because “creative works” that are not protected by copyright (i.e., are not human authorship) cannot reasonably be held to serve the purpose of copyright, fair use should be foreclosed as a defense of the generative machine.

In response, we will see CoP and defendants argue that because the product is already being used by professional creators, products like Udio or Suno serve both copyright and non-copyright purposes. While plausible, this defense is where I believe the courts may find the GAI’s purpose to be only “slightly transformative.” This is because the dominant purpose—indeed the only ROI available to the developers—is one that primarily does not fulfill the purpose of copyright and which, in fact, serves as a substitute for works that do serve the purpose of copyright.

Further, the consideration of GAI as a tool for creators in furtherance of copyright’s purpose runs headlong into the nascent doctrine as to how and how much use of GAI results in a protectable work. That question is a case-by-case consideration at a granular level. One musician’s use of Udio may produce a copyrightable composition and/or sound recording, while another’s use of the same product in a slightly different manner may have the opposite result. Considering the uncertainty of these hypotheticals to come, it is hard to imagine how the courts could find today that the product at issue favors a finding of transformativeness strong enough to carry the whole fair use analysis.

Chamber of Progress et al. will flood social media with anecdotal arguments, like disabled persons empowered to create thanks to GAI, or the whimsical notion that “machines learn the way people do.” These and other rationales for GAI’s value deserve specific responses, some of which I shall write. But in general, I predict these stories, like Lessig’s “children of YouTube,” will play well with some segment of the blogosphere but then, as legal arguments, will join the pile of similar fair use defenses lying on the floors of the federal courts.

Image source by

End of the Line for Controlled Digital Lending Theory

controlled digital lending

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.

The Second Circuit Court of Appeals yesterday affirmed that Internet Archive’s digital book lending model controlled digital lending (CDL) is not permitted by copyright law, including under the fair use exception. The outcome is a big deal and, at the same time, is not really big news. Copyright watchers grounded in what the law is (rather than what they wish it to be) were likely relieved but not surprised that the appeals court upheld a judgment, which the lower court originally issued just four days after oral arguments.

The lawsuit Hachette et al. v. Internet Archive was triggered by IA’s so-called National Emergency Library (NEL), an alleged response to the COVID shutdown whereby it arbitrarily made over one million digital books available without any kind of restriction. Because IA had no basis on which to claim it was responding to an “emergency” need, I maintain the NEL itself was a stunt designed to provoke litigation and take Brewster Kahle and friends’ copyright theories out for a ride. And as of yesterday, that ride is probably over.

Specifically, IA and a small group of academics hoped to prove that controlled digital lending is permitted by fair use. CDL is based on the theory that a library is permitted to make its own digital copy of a physical book in its collection and to loan either the digital or the physical copy to one reader at a time—and to then apply the scheme for as many physical copies as the library owns. Further, IA expanded the concept to create its Open Library by “partnering” with various libraries around the country to effectively count the physical copies on all those shelves and increase the number of readers to which the digital copies would be loaned.

The Second Circuit affirmed not only that the Open Library model is not permitted but that the CDL practice is not a fair use—even if it were limited to the physical books in IA’s possession. With the one distinction that the circuit court reversed the lower court’s finding that IA’s purpose is “commercial” under fair use factor one, IA lost resoundingly on all four fair use factors.

In short, CDL is not transformative under factor one because its purpose is no different than the legal eBook lending models offered by publishers. This finding informs factor four. Because the purpose of the use is not transformative, the CDL scheme amounts to a market substitute and is, therefore, a potential market harm. Factors two (nature of the works used) and three (amount of the works copied) barely warrant discussion. The books in suit are expressive in nature, which favors the plaintiff; and IA does not dispute that it copied and made whole books available. Copying whole works may be permissible, but only possibly if the use is transformative.

Since before this suit began, IA has claimed to be a champion for all libraries, even going so far as to assert that CDL is simply doing what all libraries do—i.e., to “buy books and loan books.” But that rhetoric, conveniently facile for the media, obfuscated the more complex truth that IA, its colleagues, and supporting amici were pushing a theory with no foundation in law. CDL is a self-indulgent thought experiment for a handful of academics and activists with no skin in the game of creating literary works. And IA’s claim to be acting like any other library was simply untrue.

To those who believe the law should be different to facilitate IA’s alleged social benefit, I have considered the intent of CDL (and the alleged burden of eBook licensing) and find no evidence to support the claim that CDL fulfils an unmet need. In fact, as stated in past posts, Kahle’s dream of digitizing everything and making it all freely available is not only harmful to future authorship but is a threat to local libraries. Further, it bears repeating that the publishers in this case represent thousands of authors, while IA represents the interests of no authors.

I get why even creators are often persuaded by Internet Archive’s claims. After all, creators tend to love libraries and are instinctively wary of corporations, including publishers. But to put this in contemporary context, IA’s mass digitization and “lending” scheme, which would harm future authorship, is hardly much better than AI’s mass digitization and “training” scheme, which would also harm future authorship. (See what I did with the letters there?) Mass copying and distribution of protected works should always be viewed skeptically because the purpose is rarely transformative and non-threatening to creators. As the court states in yesterday’s judgment:

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.

Udio Answers Record Labels’ Complaint in Gen AI Lawsuit

As mentioned in my last post about the record labels’ lawsuits against GAI companies Suno and Udio, I will generally focus on the latter case. Both cases are almost identical, but because UMG et al. v. Ucharted Labs Inc. is at the SDNY (in the Second Circuit), those proceedings may be followed by other courts with considerably less copyright law precedent.

Udio’s answer filed on August 1 relies substantially on the premise that there is no cause of action whatsoever. The developer intends to show that “This lawsuit…seeks a genuinely unprecedented result: a ruling that it is actionable copyright infringement, not fair use, to have copied Plaintiffs’ works as part of the process of developing a new technology, even though the ultimate outputs of that new technology are themselves non-infringing.”

The list of counter-factual evidence Udio forecasts is too long to summarize, but the heart of its legal argument at this juncture is that 1) statute explicitly bars protection of musical style; 2) its AI training process entailed learning about music rather than copying protected sound recordings; and 3) because the product’s outputs are largely (or entirely) non-infringing, the purpose of producing “new” music is “what copyright law is designed to encourage, not prohibit.” The defendant also alleges that the recording industry’s claim is invalid on the basis that it has “misused copyright law” as part of a longstanding tradition of stifling competition. So, we have a ballgame that’s going into extra innings, and there will be plenty to say about the details as they emerge.

For now, I take issue with a few premises inherent to Udio’s answer. The first, which I have already stated a few times, is that even if every song output by Udio is “new” as a matter of law, the fact that none of these outputs is a work of “new authorship” as a matter of law militates against Udio’s implication that its product expands the purpose of copyright in general. And as stated, if that is correct, this should militate against a finding of fair use.

Second, despite the fact that Udio can and does point to uses of its product by creators who are plausibly engaged in authorship is, at best, a difficult basis on which to argue that the primary purpose of the product advances authorship. For one thing, the business model appears to be based largely on providing a music toy for consumers, not a tool for creators. Next, even where Udio may be used by professional music creators, the extent to which this fosters new authorship is a case-by-case consideration—one that relies on still-developing doctrine around the use of AI and authorship.

Third, even if Udio could prove allegations of relevant, anti-competitive practices among all the record labels (and I do not mean to suggest they can), the court must remain focused on the interests of individual creators—especially the next generation of music makers. The labels’ argument that the outputs compete with demand for existing sound recordings could be read as protectionism of existing catalogs but should be considered as to whether Udio competes with, or even obviates, the need for new human authorship in music. If so, this is categorically not what copyright law is designed to foster.

As stated in a few posts, and in comments to the Copyright Office, the unique challenge presented by GAI is that rather than pose a threat to the interests of specific authors’ works, it poses a potential threat to authorship itself. In this light, Professor Jane Ginsburg, in a new paper about the state of fair use jurisprudence, discusses two points that stand out for me at the moment. First, she describes the nature of a use-based fair use analysis (as applied in Warhol), which should not “untether” the fair use protection for a use other than the one narrowly ruled on by the court. Second, she notes that the courts may look beyond the “explicit direction” of the fair use statute to consider a factor like broad effect on authors’ careers—or even the potential for other unlawful uses like forgery or fraud.

With regard to use-based analysis, Ginsburg forecasts the uncertainty in adopting a per se fair use rule for machine learning because the consideration of fair use of the inputs may turn on the nature of the outputs. “If an AI system ingests multiple images of apples, including Cézanne’s depictions (let’s assume Cézanne’s works were still under copyright) its training data will enable the system to “know” both what an apple looks like, and what a Cézanne apple looks like. The fair use inquiry may depend on whether the user asks for an apple, or for a Cézanne apple,” Ginsburg writes.

Perhaps more directly applicable to the labels’ case against Udio, Ginsburg states in regard to image-generating AIs and fair use factor four, the effect of the use on market value:

… even under a solely work-based interpretation of section 107(4), one may observe that the wholesale copying of an artist’s works into training data in order to enable stylistically similar outputs jeopardizes not only the artist’s future employment or commissions, but also devalues the actual works copied, because the image-generation program can produce outputs that compete with already-created works as well.

That same rationale would seem to apply to the labels’ evidence that Udio can output sounds which are substantially similar to famous and protected sound recordings. So, while the defendant is correct to say that copyright does not protect style and that music production relies substantially on mixing and matching a finite combination of styles, arrangements, etc., that premise, both statutory and judicial, is derived from a copyright history that has only ever included human artists in “competition” with one another. Consequently, the courts have latitude to find that it is in fact the AI developer who is seeking the novel conclusion that its machine furthers the purpose of copyright law.

As I say, there will be plenty of details to follow and plenty of considerations to nerd out on, if one is so inclined. And for better or worse, I am so inclined. Stay tuned (pun intended).