Earlier this week, the Copyright Office convened a listening session on the topic of copyright issues in AI-generated music and sound recordings, the fourth in its listening session series on copyright issues in different types of AI-generated creative works. Authors Alliance participated in the first listening session on AI-generated textual works, and we wrote about the second listening session on AI-generated images here. The AI-generated music listening session participants included music industry trade organizations like the Recording Industry Association of America, Songwriters of North America, and the National Music Publishers’ Association; generative AI music companies like Boomy, Tuney, and Infinite Album; music labels like the Universal Music Group and Wixen; and individual musicians, artists, and songwriters. Streaming service Spotify and collective-licensing group SoundExchange also participated. 

Generative AI Tools in the Music Industry

Many listening session participants discussed the fact that some musical artists, such as Radiohead and Brian Eno, have been using generative AI tools as part of their work for decades. For those creators, generative AI music is nothing new, but rather an expansion of existing tools and techniques. What is new is the ease with which ordinary internet users without musical training can assemble songs using AI tools—programs like Boomy enable users to generate melodies and musical compositions, with options to overlay voices or add other sounds. Some participants sought to distinguish generative tools from so-called “assistive tools,” with the latter being more established for professional and amateur musicians. 

Where some established artists themselves have long relied on assistive AI tools to create their works, AI-generated music has lowered barriers to entry for music creation significantly. Some take the view that this is a good thing, enabling more creation by more people who could not otherwise produce music. Others protest that those with musical talent and training are being harmed by the influx of new participants in music creation, as these types of songs flood the market. In my view, it’s important to remember that the purpose of copyright, furthering the progress of science and the useful arts, is served when more people can generate creative works, including music. Yet AI-generated music may already be at or past the point where it can be indistinguishable from works created by human artists without the use of these tools, at least to some listeners. It may be the case that, as at least one participant suggested, audio generated works are somehow different from AI-generated textual works such that they may require different forms of regulation. 

Right of Publicity and Name, Image, and Likeness

Although the topic of the listening session was federal copyright law, several participants discussed artists’ rights in both their identities and voices—aspects of the “right of publicity” or the related name, image, and likeness (“NIL”) doctrine. These rights are creatures of state law, rather than federal law, and allow individuals, particularly celebrities, to control what uses various aspects of their identities may be put to. In one well-known right of publicity case, Ford used a Bette Midler “sound alike” for a car commercial, which was found to violate her right of publicity. That case and others like it have popularized the idea that the right of publicity can cover voice. This is a particularly salient issue within the context of AI-generated music due to the rise of “soundalikes” or “voice cloning” songs that have garnered substantial popularity and controversy, such as the recent Drake soundalike, “Heart on My Sleeve.” Some worry that listeners could believe they are listening to the named musical artist when in fact they are listening to an imitation, potentially harming the market for that artist’s work. 

The representative from the Music Artists Coalition argued that the hodge podge of state laws governing the right of publicity could be one reason why soundalikes have proliferated: different states have different levels of protection, and the lack of unified guidance on how these types of songs are governed under the law can create uncertainty as to how they will be regulated. And the representative from Controlla argued that copyright protection should be expanded to cover voice or identity rights. In my view, expanding the scope of copyright in this way is neither reasonable nor necessary as a matter of policy (and furthermore, would be a matter for Congress, and not the Copyright Office, to address), but it does show the breadth of the soundalike problem for the music industry. 

Copyrightability of AI-Generated Songs

Several listening session participants argued for intellectual property rights in AI-generated songs, and others argued that the law should continue to center human creators. The Copyright Office’s recent guidance regarding copyright in AI-generated works suggests that the Office does not believe that there is any copyright in the AI-generated materials due to the lack of human authorship, but human selection, editing, and compilation can be protected. The representatives from companies with AI-generating tools expressed a need for some form of copyright protection for the songs these programs produce, explaining that they cannot be effectively commercialized if they are not protected. In my view, this can be accomplished through protection for the songs as compilations of uncopyrightable materials or as original works, owing to human input and editing. Yet, as many listening session participants across these sessions have argued, the Copyright Office registration guidance does not make clear precisely how much human input or editing is needed to render an AI-generated work a protectable original work of authorship. 

Licensing or Fair Use of AI Training Data

In contrast to the view taken by many during the AI-generated text listening session, none of the participants in this listening session argued outright that training generative AI programs on in-copyright musical works was fair use. Instead, much of the discussion focused on the need for a licensing scheme for audio materials used to train generative AI audio programs. Unlike the situations with many text and image-based generative AI programs, the representatives from generative AI music programs expressed an interest and willingness to enter into licensing agreements with music labels or artists. In fact, there is some evidence that licensing conversations are already taking place. 

The lack of fair use arguments during this listening session may be due to the particular participants, industry norms, or the “safety” of expressing this view in the context of the music industry. But regardless, it provides an interesting contrast to views around training data text-generating programs like ChatGPT, which many (including Authors Alliance) have argued are fair uses. This is particularly remarkable since at least some of these programs, in our view, use the audio data they are trained on for a highly transformative purpose. Infinite Album, for example, allows users to generate “infinite music” to accompany video games. The music reacts to events in the video game—becoming more joyful and upbeat for victories, or sad for defeats—and can even work interactively for those streaming their games, where those watching the stream can temporarily influence the music. This seems like precisely the sort of “new and different purpose” that fair use contemplates, and similarly like a service that is unlikely to compete directly with individual songs and records. 

Generative AI and Warhol Foundation v. Goldsmith

Many listening session participants discussed the interactions between how AI-generated music should be regulated under copyright law and the recent Supreme Court fair use decision in Warhol Foundation v. Goldsmith (you can read our coverage of that decision here), which also considered whether a particular use which could have been licensed was fair use. And some participants argued that the decision in Goldsmith makes it clear that training generative AI models (i.e., the input stage) is not a fair use under the law. It is not clear precisely how the decision will impact the fair use doctrine going forward, particularly as it applies to generative AI, and I think it is a stretch to call it a death knell for the argument that training generative AI models is a fair use. However, the Court did put a striking emphasis on the commerciality of the use in that case, deemphasizing the transformativeness inquiry somewhat. This could impact the fair use inquiry in the context of generative AI programs, as these programs tend overwhelmingly to be commercial, and the outputs they create can and are being used for commercial purposes. 

Yesterday, the Supreme Court released two important decisions in Warhol Foundation v. Goldsmith and Gonzalez v. Google—cases that Authors Alliance has been deeply invested in, submitting amicus briefs to the Court in both cases. 

Warhol Foundation v. Goldsmith and Transformativeness

First, the Court issued its long-awaited opinion in Warhol Foundation v. Goldsmith, a case Authors Alliance has been following for years, and for which we submitted an amicus brief last summer. The case concerned a series of screen prints of the late musical artist Prince created by Andy Warhol, and asked whether the creation and licensing of one of the images, an orange screen print inspired by Goldsmith’s black and white photograph (which the Court calls “Orange Prince”), constituted fair use. After the Southern District of New York found for the Warhol Foundation on fair use grounds, the Second Circuit overturned the ruling, finding that the Warhol Foundation’s use constituted infringement. The sole question before the Supreme Court was whether the first factor in fair use analysis, the purpose and character of the use, favored a finding of fair use. 

To our disappointment, the Supreme Court’s majority agreed with the holding of the district court, finding that the purpose and character of Warhol’s use favored Goldsmith, such that it did not support a finding of fair use. This being said, the decision focused narrowly on the Warhol Foundation’s “commercial licensing of Orange Prince to Condé Nast,” expressing “no opinion as to the creation, display, or sale of any of the original Prince Series works.” Because the Court cabins its opinion, focusing specifically on the licensing of Orange Prince to Condé Nast rather than the creation of the entire Prince series, the decision is less likely to have a deleterious effect on the fair use doctrine generally than a broader decision would have. 

Writing for the majority, Justice Sotomayor argued that Goldsmith’s photograph and the Prince screen print in question shared the same purpose, “portraits of Prince used to depict Prince in magazine stories about Prince.” Moreover, the Court found the use to be commercial, given that the screen print was licensed to Condé Nast. Justice Sotomayor explained that “if an original work and secondary use share the same or highly similar purposes, and the secondary use is commercial, the first fair use factor is likely to weigh against fair use, absent some other justification for copying.” Justice Sotomayor found that the two works shared the same commercial purpose, and therefore concluded that factor one favored Goldsmith. 

Justice Kagan, joined by Chief Justice Roberts, issued a strongly worded dissenting opinion. The dissent admonished the majority for its departure from Campbell’s “new meaning or message test,” an inquiry that Authors Alliance advocated for in our amicus brief. Justice Kagan further criticized the majority’s shifting focus towards commerciality, arguing that the fact that the use was a licensing transaction should not be given so much importance in the analysis. While Authors Alliance agrees with these points, we are less sure that the majority’s decision goes so far as to “constrain creative expression” or “threaten[] the creative process. And while it’s uncertain what effect this case will have on the fair use doctrine more generally, one important takeaway is that the question of whether the use in question is commercial in nature—a consideration under the first factor—has been elevated to one of greater importance. 

While we thought this case offered a good opportunity for the Court to affirm a more nuanced approach to transformative use, we much prefer the Supreme Court’s approach to the Second Circuit’s decision, and applaud the Court on confining its ruling to the narrow question at issue. The holding does not, in our view, radically alter the doctrine of fair use or disrupt a bulk of established case law. Moreover, some aspects of arguments we made in our brief—such as the notion that transformativeness is a matter of degree, not a binary—are present in the Court’s decision. This is a good thing, in our view, as it will allow for more nuanced consideration of a use’s character and purpose, and stands in contrast to the Second Circuit’s all or nothing view of transformativeness. 

Gonzalez v. Google and the Missing Section 230

Also yesterday, the Court released its opinion in Gonzalez v. Google, a case that generated much attention because of its potential threat to Section 230, and another case in which Authors Alliance submitted an amicus brief. The case asked whether Google could be held liable under an anti-terrorism statute for harm caused by ISIS recruitment videos that YouTube’s algorithm recommended. In its per curiam decision (a unanimous one without a named Justice as author), the Court stated that Gonzalez’s complaint had failed to state a viable claim under the relevant anti-terrorism statute. Therefore, it did not reach the question of the applicability of Section 230 to the recommendations at issue. In other words, a case that generated tremendous concern about the Court disturbing Section 230 and harming internet creators, communities, and services that relied on it ended up saying nothing at all about the statute. 

Yesterday, I represented Authors Alliance in a Copyright Office listening session on copyright issues in AI-generated literary works, in the first of two of such sessions that the Office convened yesterday afternoon. I was pleased to be invited to share our views with the Office and participate in a rousing discussion among nine other stakeholders, representing a diverse group of industries and positions. Generative AI raises challenging legal questions, particularly for its skeptics, but it also presents some incredible opportunities for authors and other creators.

During the listening session, I emphasized the potential for generative AI programs (like OpenAI’s Chat GPT, Microsoft’s Bing AI, Jasper, and others) to support authorship in a number of different ways. For instance, generative AI programs support authors by increasing the efficiency of some of the practical aspects of being a working author aside from their writings. But more importantly, generative AI programs can actually help authors express themselves and create new works of authorship. 

In the first category, generative AI programs can support authors by, for example, helping them create text for pitch letters to send to agents and editors, produce copy for their professional websites, and develop marketing strategies for their books. Making these activities more efficient frees up time for authors to focus on their writing, particularly for authors whose writing time is limited by other commitments. 

In the second category, generative AI has tremendous potential to help authors come up with new ideas for stories, develop characters, summarize their writings, and perform early stage edits of manuscripts. Moreover, and particularly for academic authors, generative AI can be an effective research tool for authors seeking to learn from a large corpus of texts. Generative AI programs can help authors research by providing short and simple summaries of complex issues, surveys of the landscape of various fields, or even guidance on what human works to turn to in their research. Authors Alliance is committed to protecting authors’ right to conduct research, and we see generative AI tools as a new, innovative, and efficient form of conducting this research. Making research easier helps authors save time, and has a particular benefit for authors with disabilities that make it difficult to travel to multiple libraries or otherwise rely on analog forms of research. 

These programs undoubtedly have the potential to serve as powerful creative tools that support authorship in these ways and more, but, when discussing the copyright implications of the programs and the works they produce, it’s important to remember just how new these technologies are. Because generative AI remains in its infancy, and the costs and benefits for different segments of the creative industry have yet to be seen, it seems to me to be sensible to preserve the development of these tools before crafting legal solutions to problems they might pose in the future. And in fact, in our view, U.S. copyright law already has the tools to deal with many of the legal challenges that these programs might post. When generative AI outputs look too much like the copyrighted inputs they are trained on, the substantial similarity test can be used to assess claims of copyright infringement to vindicate an authors’ exclusive rights in their works when those outputs do infringe. 

In any case, in order for generative AI programs to be effective creative tools, it’s necessary that they are trained on large corpora. Narrowing the corpus of works the programs are trained on—through compulsory licensing or other mechanisms—can have disastrous effects. For example, research has shown that narrow data sets are more likely to produce racial and gender bias in AI outputs. In our view, the “input” step, where the programs are trained on a large corpus of works, is a fair use of these texts. And the holdings in Google Books and HathiTrust indicate that it is consistent with fair use to build large corpora of works, including works that remain protected by copyright, for applications such as computational research and information discovery. Additionally, the Copyright Office has recognized this principle in the context of research and scholarship, as demonstrated by its approval of Authors Alliance’s petition for an exemption from DMCA restrictions for text and data mining. 

The question of the copyright status of AI-generated works is an important one. Most if not all of the stakeholders participating in this discussion agreed with the Copyright Office’s recent guidance regarding registration in AI-generated works: under ordinary copyright principles, the lack of human authorship means these texts are not protected by copyright. This being said, we also recognize that there may be challenges in reconciling existing copyright principles with these new types of works and the questions about authorship, creativity, and market competition that they might pose. 

But importantly, while this technology is still in its early stages, it serves the core purposes of copyright—furthering the progress of science and the useful arts by incentivizing new creation—to allow these systems to develop and confront new legal challenges as they emerge. Copyright is not only about protecting the exclusive rights of copyright holders (a concern that underlies many arguments against generative AI as a fair use), but incentivizing creativity for the public benefit. The new forms of creation made possible through generative AI can incentivize people who would not otherwise create expressive works to do so, bringing more people into creative industries and adding new creative expression to the world to the benefit of the public.

The listening sessions were recorded, and will be available on the Copyright Office website in the coming weeks. And these listening sessions are only the beginning of the Office’s investigation of copyright in AI generated works. Other listening sessions on visual works, music, and audiovisual works will be held in the coming weeks, and the Office has indicated that there will be an opportunity for written public comments in order for stakeholders to weigh in further. We are committed to remaining involved in these cutting edge issues, through written comments and otherwise, and we will keep our readers informed as policy around generative AI continues to evolve. 

Yesterday, Authors Alliance submitted a comment to the U.S. Copyright Office in response to a notice of proposed rulemaking asking for feedback from the public on new rules to govern ex parte communications. “Ex parte communications” refer to communications outside the normal, permitted channels of communication—in this case, to communications between organizations or members of the public and Copyright Office staff outside of hearings or other formal proceedings. Ex parte communications with the Copyright Office are important, because they allow stakeholders and the office to work out open questions in rulemakings or other proceedings outside of the formal channels. Authors Alliance relied on our ability to make ex parte communications during the last Section1201 rulemaking cycle (where we obtained our text data mining exemption) in order to clarify certain issues. Now, the Office is proposing establishing formal rules for how these communications can be made, as well as establishing transparency around them. We support this proposal, and shared our thoughts in a comment. You can read our full comment here.

On Friday, Southern District of New York Judge John Koeltl issued a much-anticipated decision in Hachette Books v. Internet Archive. Unfortunately, as many of our members and allies are aware, the judge ruled against the Internet Archive, finding that its CDL program was not protected by the doctrine of fair use and granting the publishers’ motion for summary judgment. You can read the 47-page decision for yourself here. 

In his fair use analysis, Judge Koeltl found that each of the four fair use factors weighed in favor of the publishers, emphasizing above all else his view that IA’s controlled digital lending program was not transformative, an important consideration under the first fair use factor, which considers the purpose and character of the use. This inquiry also involves asking whether the use in question was commercial. To the surprise of many, the decision stated that IA’s use of the publishers’ works was commercial, because the Open Library is part of the IA’s website, which it uses “to attract new members, solicit donations, and bolster its standing in the library community.” The judge found this to be the case in spite of the fact that IA “does not make a monetary profit” from CDL. In other words, the judge held that the indirect, attenuated benefits the Internet Archive (which is, after all, a nonprofit) reaps from operating the Open Library makes its CDL program commercial. 

Judge Koeltl gave less attention to the fourth factor in the fair use analysis, “the effect of the use on the potential market for the work,” which is often held up to be of significant importance. One consideration under this factor is whether the use creates a competing substitute with the original work. Unfortunately, on this point too, the court—in our view—missed the mark. This is because the decision does not draw a distinction between CDL scans and ebooks, going so far as to call CDL scans “ebooks” throughout. As we explained in our summary of the proceedings last week, many features of both CDL and ebooks make them both functionally and aesthetically distinct from one another. By glossing over these differences, the judge reached the conclusion that CDL scans are direct substitutes for licensed ebooks.

Authors Alliance is deeply concerned about the ramifications of this decision, which was exceedingly broad in scope, striking a tremendous blow to the CDL model, rather than only IA’s implementation of it. Local libraries across the country practice CDL, and library patrons and authors alike depend on it to read, research, and participate in academic discourse. 

As it stands, this decision only applies to Internet Archive and is only about the 127 books on which the publishers based their lawsuit. It does not set a binding precedent for any other library, but if left in place (or worse, if affirmed on appeal), it could cause libraries to avoid digitizing and lending books under a CDL model, which in our view would not serve the interests of many authors. This decision makes it harder for those authors to reach wide audiences: CDL enables many authors to reach more readers than they could otherwise, and authors like our members who write to be read would not be served if fewer readers could access their books. 

The decision also hampers efforts to preserve books—aside from IA’s scanning program, there are few if any centralized efforts to preserve books in digital format once their commercial life is over. Without CDL, those books could quite literally disappear, and the knowledge they advance could be lost. IA’s scanning operations do preserve such books, which is one reason we have strongly supported them in this lawsuit. By the same token, if this decision stands, it will also limit authors’ ability to conduct efficient research online. The CDL survey we launched last year revealed that CDL is an effective research tool for authors who need to consult other books as part of their writing process, and in many cases it enables them to access far more works than they could at their local library alone. Authors who rely on CDL in this way would be harmed by this decision, as they could well be forced to undergo a more time-consuming research process, detracting from time that could be spent writing. 

The Internet Archive has already indicated that it will be appealing Judge Koeltl’s ruling, and we look forward to supporting those efforts. We will continue to keep our readers and members apprised of updates as this case moves forward.

Earlier today, Judge John Koeltl of the Southern District of New York heard oral arguments in Hachette Book Group v. Internet Archive—a case Authors Alliance has been following since the lawsuit was first filed back in 2020. The case is about—among other things—whether Internet Archive’s controlled digital lending program qualifies as a fair use. Authors Alliance submitted an amicus brief in support of the Internet Archive back in July, arguing that CDL serves the interests of authors who write to be read. IA’s attorney cited to our brief during oral argument, and we are pleased that we were able to magnify the voices of authors who write to be read through its submission. You can learn more about the case and read our brief here.

In the hearing, the judge considered each party’s motion for summary judgment. The parties hotly contested a number of key issues in the case, including whether each side’s experts had properly demonstrated market harm (or lackthereof), what the appropriate market to consider was for purposes of fair use analysis, the commerciality of IA’s use, and what legal cases supported both arguments in favor of and against fair use. Judge Koeltl asked the Internet Archive’s attorney a number of probing questions on these points, grappling with the difficult questions in this case. The judge further implied that there may be open issues of fact in this case, which could indicate the need for additional briefings or hearings. 

CDL and Commerciality

The parties disagreed on the commerciality of IA’s use when it produces and makes CDL scans available. The publishers attorney argued that IA’s CDL operations are “intertwined” with its other functions, such as its ownership of the book vendor Better World Books, and further emphasizing its argument that CDL loans result in lost revenue for the publisher—in other words, that the supposed commercial harm to the publishers that results from CDL lending makes the CDL lending itself commercial. The Internet Archive’s attorney answered that IA is a nonprofit organization that does not profit at all from its CDL program. He pointed to the fact that traditional library lending is not commercial in nature and does not provide libraries like IA with commercial benefits. 

CDL and Market Effects

The plaintiffs’ attorney began by setting forth plaintiffs’ views on the issue of market harm—the fourth factor in fair use analysis, often cited as one of the most important factors in the inquiry. Plaintiffs discussed what they see as massive financial harm stemming from IA’s CDL program, which they estimated to amount to “millions of dollars in licensing revenues.” Plaintiffs also emphasized that, were CDL “given the green light,” or upheld as a fair use, the plaintiffs would suffer even greater losses. Throughout her argument, plaintiffs’ attorney emphasized the “basic economic principle and common sense is that you cannot compete with free.” In other words, the publishers argue that the ebook library licensing market could collapse altogether if CDL were allowed to continue. Yet this misses the point that CDL is a longstanding and established practice, which has seen adoption and growth in libraries across the country while the ebook licensing market has continued to thrive. 

Judge Koeltl, however, pressed the publishers on whether they had shown evidence of actual market harm, i.e. proof that IA’s CDL program had directly harmed their bottom line. In response, plaintiffs criticized the expert evidence offered by IA’s experts to show that no such harm had occurred. This is a difficult question because the party asserting a fair use defense typically has the burden of showing that the use has not harmed the market, but it exceedingly difficult to prove a negative. 

The judge also questioned whether CDL actually could represent such a loss: the publishers’ argument rests on the premise that libraries loan out CDL scans in lieu of paying to license ebooks, and were CDL not permitted under the law, IA and other libraries would instead choose to pay licensing fees to lend out ebooks. The judge pointed out that the result might in fact be that libraries would choose not to lend digital copies of works out at all, or would instead lend out physical books, undercutting the lost licensing revenue argument. 

IA’s attorney argued that the publishers had not offered empirical evidence of market harm in this case, focusing on the fact that when a library lends out a CDL scan, it does so in lieu of a physical book, “simulating the limitations of physical books.” This is due to CDL’s “owned to loaned” ratio requirement: a library can only loan out the number of CDL scans as it has physical books in its collection, and can only loan these scans out to one patron at a time. When a library lends out a CDL scan, it does so in lieu of loaning the physical book, for which it has already paid. And while the plaintiffs mentioned harm to authors (who are, after all, the people that copyright law is intended to protect) several times during their argument, they did this in a way that linked authors with publishers as parties that are financially invested in a works’ sale—author interests and the finer details of the economics of author income and library lending were absent from the discussion. 

The parties also disagreed about which market was the appropriate one to look to when discussing market harm in the context of fair use analysis. The publishers argued, and the judge seemed to assume, that the proper market is the library ebook licensing market. The judge opined that libraries could, instead of using CDL to lend out their books, simply purchase an ebook license. He seemed to view CDL scans and licensed ebooks as one and the same, despite the fact that there are several key differences between these types of loans, both in form and function, as explained in other amicus briefs in the case. Moreover, missing from the argument was the fact that, in many cases, libraries loan out CDL scans because no ebook is available to them: particularly for older books in a publisher’s backlist, or for books that are no longer available commercially, there is in many cases no ebook available, or no ebook available to libraries. Library patrons with print or mobility disabilities in need of digital copies of these kinds of works in order to read them would be greatly harmed if CDL were no longer permitted. 

CDL and Transformativeness

The publishers’ attorney started from the premise that CDL as a use was not transformative, explaining that a licensed ebook and a CDL scan served precisely the same function. In response, IA’s attorney in response argued that CDL is a transformative use because it “utilizes technology to achieve the transformative purpose of improving efficiency of delivering content without unreasonably encroaching on the rights of the rightsholder.” He further explained that fair uses are favored when they serve the key purpose of copyright: incentivizing new creation for the public benefit without harming the interests of rightsholders. To illustrate these benefits, he cited to Authors Alliance’s amicus brief, in which we explained the myriad ways that CDL benefits authors and can even incentivize the creation of new works. 

Adding to its transformativeness argument, IA explained that, when it comes to speculative or actual market harm, such an effect must be balanced against the public benefit that results from the use. And when it comes to CDL, this public benefit is tremendous: numerous amici, as well as Authors Alliance, explained that CDL serves the interests of library patrons, authors, and the public writ large. 

What’s Next?

Now that the judge has heard both sides’ arguments, he will issue a decision in the case. While there is no way of knowing exactly when this will happen, Judge Koeltl is known for issuing decisions fairly quickly, so we may have a decision as soon as later this week. As always, we will keep our members and readers apprised of any developments in this pivotal case as it moves forward.