Author Archives: Rachel Brooke

Copyright and Generative AI: Our Views Today

Posted August 30, 2023
Large copyright sign made of jigsaw puzzle pieces” by Horia Varlan is licensed under CC BY 2.0.

Authors Alliance readers will surely have noticed that we have been writing a lot about generative AI and copyright lately. Since the Copyright Office issued its decision letter on copyright registration in a graphic novel that included AI-generated images a few months back, many in the copyright community and beyond have struggled with the open questions around generative AI and copyright.

The Copyright Office has launched an initiative to study generative AI and copyright, and today issued a notice of inquiry to solicit input on the issues involved. The Senate Judiciary Committee has also held multiple hearings on IP rights in AI-generated works, including one last month focused on copyright. And of course there are numerous lawsuits pending over its legality, based on theories ranging from copyright infringement to to privacy to defamation. It’s also clear that there is little agreement about a one-size-fits-all rule for AI-generated works that applies across industries. 

At Authors Alliance, we care deeply about access to knowledge because it supports free inquiry and learning, and we are enthusiastic about ways that generative AI can meaningfully further those ideals. In addition to all the mundane but important efficiency gains generative AI can assist with, we’ve already seen authors incorporate generative AI into their creative processes to produce new works. We’ve also seen researchers incorporate these tools to help make new discoveries. There are some clear concerns about how generative AI tools, for example, can make it easier to engage in fraud and deception, as well as perpetuating disinformation. There have been many calls for legal regulation of generative AI technologies in recent months, and we wanted to share our views on the copyright questions generative AI poses, recognizing that this is a still-evolving set of questions.  

Copyright and AI

Copyright is at its core an economic regulation meant to provide incentives for creators to produce and disseminate new expressive works. Ultimately, its goal is to benefit the public by promoting the “progress of science,” as the U.S. Constitution puts it. Because of this, we think new technology should typically be judged by what it accomplishes with respect to those goals, and not by the incidental mechanical or technological means that it uses to achieve its ends. 

Within that context, we see generative AI as raising three separate and distinct legal questions. The first and perhaps most contentious is whether fair use should permit use of copyrighted works as training data for generative AI models. The second is how to treat generative AI outputs that are substantially similar to existing copyrighted works used as inputs for training data—in other words, how to navigate claims that generative AI outputs infringe copyright in existing works. The third question is whether copyright protection should apply to new outputs created by generative AI systems. It is important to consider these questions separately, and avoid the temptation to collapse them into a single inquiry, as different copyright principles are involved. In our view, existing law and precedent give us good answers to all three questions, though we know those answers may be unpalatable to different segments of a variety of content industries. 

Training Data and Fair Use

The first area of difficulty concerns the input stage of generative AI. Is the use of training data which includes copyrighted works a fair use, or does it infringe on a copyright owner’s exclusive rights in her work? The generative AI models used by companies like OpenAI, Stability AI, and Stable Diffusion are based on massive sets of training data. Much of the controversy around intellectual property and generative AI concerns the fact that these companies often do not seek permission from rights holders before training their models on works controlled by these rights holders (although some companies, like Adobe, are building generative AI models based on their own stock images, openly-licensed images, and public domain content). Furthermore, due to the size of the data sets and nature of their collection (often obtained via scraping websites), the companies that deploy these models do not make clear what works make up the training data. This question is one that is controversial and highly debated in the context of written works, images, and songs. Some creators and creator communities in these areas have made calls for “consent, credit, and compensation” when their works are included in training data. The obstacle to that point of view is, if the use of training data is a fair use, none of this is required, at least not by copyright.  

We believe that the use of copyrighted works as training data for generative AI tools should generally be considered fair use. We base this view on our reading of numerous fair use precedents including Google Books and HathiTrust cases as well others such as iParadigms. These and other cases support the idea that fair use allows for copying for non-expressive uses—copying done as an “intermediate step” in producing non-infringing content, such as by extracting non-expressive content such as patterns, facts, and data in or about the work. The notion that non-expressive (also called “non-consumptive”) uses do not infringe copyrights is based in large part on a foundational principle in copyright law: copyright protection does not extend to facts or ideas. If it did, copyright law would run the risk of limiting free expression and inhibiting the progress of knowledge rather than furthering it. Using in-copyright works to create a tool or model with a new and different purpose from the works themselves, which does not compete with those works in any meaningful way, is a prototypical fair use. Like the Google Books project (as well as text data mining), generative AI models use data (like copyrighted works) to produce information about the works they ingest, including abstractions and metadata, rather than replicating expressive text. 

In addition, fair use of copyrighted works as training data for generative AI has several practical implications for the public utility of these tools. For example, without it, AI could be trained on only “safe materials,” like public domain works or materials specifically authorized for such use. Models already contain certain filters—often excluding hateful content or pornography as part of its training set. However, a more general limit on copyrighted content—virtually all creative content published in the last one hundred years—would tend to amplify bias and the views of an unrepresentative set of creators. 

Generative AI Outputs and Copyright Infringement

The feature that most distinguishes generative AI from technology in copyright cases that preceded it, such as Google Books and HathiTrust, is that generative AI not only ingests copyrighted works for the purpose of extracting data for analysis or search functionality, but for using this extracted data to produce new content. Can content produced by a generative AI tool infringe on existing copyrights?

Some have argued that the use of training data in this context is not a fair use, and is not truly a “non-expressive use” because generative AI tools produce new works based on data from originals and because these new works could in theory serve as market competitors for works they are trained on. While it is a fair point that generative AI is markedly different from those earlier technologies because of these outputs, the point also conflates the question of inputs and outputs. In our view, e using copyrighted works as inputs to develop a generative AI tool is generally not infringement, but this does not mean that the tool’s outputs can’t infringe existing copyrights. 

We believe that while inputs as training data is largely justifiable as fair use, it is entirely possible that certain outputs may cross the line into infringement. In some cases, a generative AI tool can fall into the trap of memorizing inputs such that it produces outputs that are essentially identical to a given input. While evidence to date indicates that memorization is rare, it does exist

So how should copyright law address outputs that are essentially memorized copies of inputs? We think the law already has the tools it needs to address this. Where fair use does not apply, copyright’s “substantial similarity” doctrine is equipped to handle the question of whether a given output is similar enough to an input to be infringing. The substantial similarity doctrine is appropriately focused on protection of creative expression while also providing room for creative new uses that draw on unprotectable facts or ideas. Substantial similarity is nothing new: it has been a part of copyright infringement analysis for decades, and is used by federal courts across the country. And it may well be that standards, such as a set of  “Best Practices for Copyright Safety for Generative AI” proposed by law professor Matthew Sag, will become an important measure of assessing whether companies offering generative AI have done enough to guard against the risk of their tools producing infringing outputs.

Copyright Protection of AI Outputs

A third major question is, what exactly is the copyright status of the outputs of generative AI programs: are they protected by copyright at all, and if so, who owns those copyrights? Under the Copyright Office’s recent registration guidance, the answer seems to be that there is no copyright protection in the outputs. This does not sit well with some generative AI companies or many creators who rely on generative AI programs in their own creative work. 

We generally agree with the Copyright Office’s recent guidance concerning the copyright status of AI-generated works, and believe that they are unprotected by copyright. This is based on the simple but enduring “human authorship” requirement in copyright law, which dates back to the late 19th century. In order to be protected by copyright, a work must be the product of a human author and contain a modicum of human creativity. Purely mechanical processes that occur without meaningful human creative input cannot generate copyrightable works. The Office has categorized generative AI models as this kind of mechanical tool: the output responds to the human prompt, but the human making the prompt does not have sufficient control over how the model works to make them an “author” of the output for the purposes of copyright law. The district court for D.C. recently issued a decision agreeing with this take in Thaler v. Perlmutter, a case that challenged the human authorship requirement in the context of generative AI. 

It’s interesting to note here that in the Copyright Office listening session on text-based works, participants nearly universally agreed that outputs should not be protected by copyright, agreeing with the Copyright Office’s guidance. Yet the other listening sessions had more of a diversity of views. In particular, the participants in the listening sessions on audiovisual works and sound recordings were concerned about this issue. In industries like the music and film industries, where earlier iterations of generative AI tools have long been popular (or are even industry norms), the prospect of being denied copyright protection in songs or films, simply due to the tools used, can understandably be terrifying for creators who want to make a profit from their works. On this front, we’re sympathetic. Creators who rely on their copyrights to defend and monetize their works should be permitted to use generative AI as a creative tool without losing that protection. While we believe that the human authorship requirement is sound, it would be helpful to have more clarity on the status of works that incorporate generative AI content. How much additional human creativity is needed to render an AI-generated work a work of human authorship, and how much can a creator use a generative AI tool as part of their creative process without foregoing copyright protection in the work they produce? The Copyright Office seems to be grappling with these questions as well and seeking to provide additional guidance, such as in a recent webinar with more in-depth registration guidance for creators relying on generative AI tools in their creative efforts.

Other Information Policy Issues Affecting Authors

Generative AI has generated questions in other areas of information policy beyond the copyright questions we discuss above. Fraudulent content or disinformation, the harm caused by deep fakes and soundalikes, defamation, and privacy violations are serious problems that ought to be addressed. Those uses do nothing to further learning, and actually pollute public discourse rather than enhance it. They can also cause real monetary and reputational harm to authors. 

In some cases, these issues can be addressed by information policy doctrines outside of copyright, and in others, they can be best handled by regulations or technical standards addressing development and use of generative AI models. A sound application of state laws such as defamation law, right of publicity laws, and various privacy torts could go a long way towards mitigating these harms. Some have proposed that the U.S. implement new legislation to enact a federal right of publicity. This would represent a major change in law and the details of such a proposal would be important. Right now, we are not convinced that this would serve creators better than the existing state laws governing the right of publicity. While it may take some time for courts to figure out how to adapt legal regimes outside of copyright to questions around generative AI, adapting the law to new technologies is nothing new. Other proposals call for regulations like labeling AI-generated content, which could also be reasonable as a tool to combat disinformation and fraudulent content. 

In other cases, creators’ interests could be protected through direct regulation of the development and use of generative AI models. For example, certain creators’ desire for consent, credit, and compensation when their works are included in training data sets for generative AI programs is an issue that could be perhaps addressed through regulation of AI models. As for consent, some have called for an opt-out system where creators could have their works removed from the training data, or the deployment of a “do not train” tag similar to the robots.txt “do not crawl” tag. As we explain above, under the view that training data is generally a fair use, this is not required by copyright law. But the views that using copyrighted training data without some sort of recognition of the original creator is unfair, which many hold, may support arguments for other regulatory or technical approaches that would encourage attribution and pathways for distributing new revenue streams to creators. 

Similarly, some have called for collective licensing legislation for copyrighted content used to train generative AI models, potentially as an amendment to the Copyright Act itself. We believe that this would not serve the creators it is designed to protect and we strongly oppose it. In addition to conflicting with the fundamental principles of fair use and copyright policy that have made the U.S. a leader in innovation and creativity, collective licensing at this scale would be logistically infeasible and ripe for abuse, and would tend to enrich established, mostly large rights holders while leaving out newer entrants. Similar efforts several years ago were proposed and rejected in the context of mass digitization based on similar concerns.  

Generative AI and Copyright Going Forward

What is clear is that the copyright framework for AI-generated works is still evolving, and just about everyone can agree on that. Like many individuals and organizations, our views may well shift as we learn more about the real-world impacts of generative AI on creative communities and industries. It’s likely that as these policy discussions continue to move forward and policymakers, advocacy groups, and the public alike grapple with the open questions involved, the answers to these open questions will continue to develop. Changes in generative AI technology and the models involved may also influence these conversations. Today, the Copyright Office published issued a notice of inquiry on the topic of copyright in AI-generated works. We plan to submit a comment sharing our perspective, and are eager to learn about the diversity of views on this important issue.

Copyright Protection in AI-Generated Works Update: Decision in Thaler v. Perlmutter

Posted August 24, 2023
Photo by Google DeepMind on Unsplash

Last week, the District Court for the District of Columbia announced a decision in Thaler v. Perlmutter, a case challenging copyright’s human authorship requirement in the context of a work produced by a generative AI program. This case is one of many lawsuits surrounding copyright issues in generative AI, and surely will not be the last we hear about the copyrightability of AI-generated works, and how this interacts with copyright’s human authorship requirement. In today’s post, we’ll provide a quick summary of the case and offer our thoughts about what this means for authors and other creators.

Background

Back in 2018 (before the current public debate about copyright and generative AI had reached the fever pitch we now see today), Dr. Stephen Thaler applied for copyright registration in a work of visual art produced by a generative AI system he created, called the Creativity Machine. Thaler sought to register his work as a computer-generated “work-made-for-hire” since he created the machine, which “autonomously” produced the work. After a lot of back and forth with the Copyright Office, it maintained its denial of the application, explaining that the human authorship requirement in copyright law foreclosed protection for the AI-generated work, since it was not the product of a human’s creativity.

Then, Thaler then sued Shira Perlmutter, the Register of Copyrights, in the D.C. district court, asking the court to decide “whether a work autonomously generated by an AI system is copyrightable.” Judge Baryl A. Howell upheld the Copyright Office’s decision, explaining that under the plain language of the Copyright Act, “an original work of authorship” required that the author be a human “based on centuries of settled understanding” and a dictionary definition of “author.” She also cited to the U.S. Constitution’s IP clause, which similarly mentions “authors and inventors,” and over a century of Supreme Court precedent to support this principle.

Thaler’s attorney has indicated that he will be appealing the ruling to the D.C. Circuit court of appeals, and it remains to be seen whether that court will affirm the ruling. 

Implications for copyright law

The headline takeaway from this ruling is that AI generated art is not copyrightable because copyright requires human authorship, which remains a requirement in copyright law. However, the ruling is actually more nuanced and contains a few subtle points worth highlighting. 

For one, this case tested not just the human authorship requirement but also the application of the work-for-hire doctrine in the context of generative AI. On one view of the issues, if Thaler created a machine capable of creating a work that would be copyrightable were it created by a human, there is a certain appeal in framing the work as one commissioned by Thaler. On this point, the court explained that since there was no copyright in the work in the first instance based on its failure to meet the human authorship requirement, this theory also did not hold water. In other words, a work-made-for-hire requires that the “hired” creator also be a human. 

It’s important to keep in mind that Thaler was in a sense testing the reach of the limited or “thin” copyright that can be granted in compilations of AI-generated work, or AI-generated work that a human has altered, thus endowing it with at least a modicum of human creativity as copyright requires. Thaler made no changes to the image produced by his Creativity Machine, and in fact, described the process to the Copyright Office as fully autonomous rather than responding to an original prompt (as is generally the case with generative AI). Thaler was not trying to get a copyright in the work in order to monetize it for his own livelihood, but—presumably—to explore the contours of copyright in computer-generated works. In other words, the case has some philosophical underpinnings (and in fact, Thaler has said in interviews that he believes his AI inventions to be sentient, a view that many of us tend to reject). But for creators using generative AI who seek to register copyrights in order to benefit from copyright protection, things are unlikely to be quite so clear-cut. And while she found the outcome to be fairly clear cut in this case, Judge Howell observed:

“The increased attenuation of human creativity from the actual generation of the final work will prompt challenging questions regarding how much human input is necessary to qualify the user of an AI system as an ‘author’ of a generated work, the scope of the protection obtained over the resultant image, how to assess the originality of AI-generated works where the systems may have been trained on unknown pre-existing works, how copyright might best be used to incentivize creative works involving AI, and more.”

What does this all mean for authors? 

For authors who want to incorporate AI-generated text or images into their own work, the situation is a bit murkier than it was for Thaler. The case itself provides little in the way of information for human authors who use generative AI tools as part of their own creative processes. But while the Copyright Office’s registration guidance tells creators what they need to do to register their copyrights, this decision provides some insight about what will hold up in court. Courts can and do overturn agency actions in some cases (in this case, the judge could have overturned the Copyright Office’s denial of Thaler’s registration application had she found it to be “arbitrary and capricious”). So the Thaler case in many ways affirms what the Copyright Office has said so far about registrability of AI-generated works, indicating that the Office is on the right track as far as their approach to copyright in AI-generated works, at least for now. 

The Copyright Office has attempted to provide more detailed guidance on copyright in “AI-assisted” works, but a lot of confusion remains. One guideline the Office promulgated in a recent webinar on copyright registration in works containing AI-generated material is for would-be registrants to disclose the contribution of an AI system when its contribution is more than “de minimis,” i.e., when the AI-generated creation would be entitled to copyright protection if it were created by a human. This means that using an AI tool to sharpen an image doesn’t require disclosure, but using an AI tool to generate one part of an image does. An author will then receive copyright protection in only their contributions to the work and the changes they made to the AI-generated portions. As Thaler shows, an author must make some changes to an AI-generated work in order to receive any copyright protection at all in that work.

All of this means, broadly speaking, that the more an author changes an AI-generated work—such as by using tools like photoshop to alter an image or by editing AI-generated text—the more likely it is that the work will be copyrightable, and, by the same token, the less “thin” any copyright protection in the work will be. While there are open questions about how much creativity is required from a human in order to transform an AI-generated work into a copyrightable work of authorship, this case has underscored that at least some creativity is required—and using an AI tool that you yourself developed to create the work does not cut it. 

The way Thaler framed his Creativity Machine as the creator of the work in question also shows that it is important to avoid anthropomorphizing AI systems—just as the court rejected the notion of an AI-generated work being a work-made-for-hire, a creative work with both generative AI and human contributions probably could not be registered as a “co-authored” work. Humans are predisposed to attribute human characteristics to non-humans, like our pets or even our cars, a phenomenon which we have seen repeatedly in the context of chat bots. Regardless, it’s important to remember that a generative AI program is a tool based on a model. And thinking of generative AI programs as creators rather than tools can distract us from the established and undisturbed principle in copyright law that only a human can be considered an author, and only a human can hold a copyright. 

Update: Consent Judgment in Hachette v. Internet Archive

Posted August 11, 2023
Photo by Markus Winkler on Unsplash

UPDATE: On Monday, August 14th, Judge Koeltl issued an order on the proposed judgement, which you can read here, and which this blog post has been updated to reflect. In his order, the judge adopted the definition of “Covered Book” suggested by the Internet Archive, limiting the permanent injunction subject to an appeal to only those books published by the four publisher plaintiffs that are available in ebook form.

After months of deadline extensions, there is finally news in Hachette Books v. Internet Archive, the case about whether Controlled Digital Lending is a fair use, which we have been covering since its inception over two years ago, and in which Authors Alliance filed an amicus brief in support of Internet Archive and CDL. On Friday, August 11th, attorneys for the Internet Archive and a group of publishers filed documents in federal court proposing “an appropriate procedure to determine the judgment to be entered in this case,” as Judge John G. Koeltl of the Southern District of New York requested

In a letter to the court, both parties indicated that they had agreed to a permanent injunction, subject to an appeal by IA, “enjoining the Internet Archive [] from distributing the ‘Covered Books’ in, from or to the United States electronically.” This means that the Internet Archive has agreed to stop distributing within the U.S. the books in its CDL collection which are published by the plaintiff publishers in the case (Hachette Book Group, HarperCollins, Wiley, and Penguin Random House), and are currently available as ebooks from those publishers. The publishers must also send IA a catalog “identifying such commercially available titles (including any updates thereto in the Plaintiffs’ discretion), or other similar form of notification,” and “once 14 days have elapsed since the receipt of such notice[,]” IA will cease distributing CDL versions of these works under the proposed judgment.

Open Questions

Last week’s proposed judgment did leave an open question, which Judge Koeltl was asked to decide before issuing a final judgment: should IA be enjoined from distributing CDL versions of books published by the four publishers where those books are available in any form, or should it only be enjoined from distributing CDL versions of these books that are available as ebooks? This difference may seem subtle, but it’s actually really meaningful. 

The publishers asked for a broader definition, whereby any of their published works that remain in print in any form are off the table when it comes to CDL. The publishers explain in a separate letter to the court that they believe that it would be consistent with the judgment to ban the IA from loaning out CDL versions of any of the commercially available books they publish, whatever the format. They argue that it should be up to the publishers whether or not to issue an ebook edition of the work, and that even when they decide not to do so (based on an author’s wishes or other considerations), IA’s digitization and distribution of CDL scans is still infringement. 

On the other hand, the Internet Archive is asked the judge to confine the injunction to books published by the four publishers that are available as ebooks, leaving it free to distribute CDL scans of the publishers’ books that are in print, but only available as print and/or audio versions. It argues that to forbid it from lending out CDL versions of books with no ebook edition available would go beyond the matters at issue in the case—the judge did not decide whether it would be a fair use to loan out CDL versions of books only available in print, because none of the works that the publishers based the suit upon were available only as print editions. Furthermore, IA explains that other courts have found that the lack of availability of a competing substitute (in this case, an ebook edition) weighs in favor of fair use under the fourth factor, which considers market competition and market harm.

It seems to me that the latter position is much more sensible. In addition to CDL scans of books only available as physical books not being at issue in the case, the fair use argument for this type of lending is quite different. One of the main thrusts of the judge’s decision in the case was his argument that CDL scans compete with ebooks, since they are similar products, but this logic does not extend to competition between CDL scans and print books. This is because the markets for digital versions of books and analogue versions of books are quite different. Some readers strongly prefer print versions of books, and some turn to electronic editions for reasons of disability, physical distance from libraries or bookstores, or simple preference. While we believe that IA’s CDL program is a fair use, its case is even stronger when it comes to CDL loans of books that are not available electronically. 

Then on Monday, August 14th, Judge Koeltl issued an order and final judgment in the case, agreeing with the Internet Archive and limiting the injunction to books published by the four publishers which are available in ebook form. Again, this may seem minor, but I actually see it as a substantial win, at least for now. While even the more limited injunction is a serious blow to IA’s controlled digital lending program, it does allow them to continue to fill a gap in available electronic editions of works. The judge’s primary reasoning was that books not available as ebooks was beyond the scope of what was at issue in the case, but he also mentioned that factor four analysis could have been different were there no ebook edition available.

Limitations of the Proposed Judgment

Importantly, the parties also stipulated that this injunction is subject to an appeal by the Internet Archive. This means that if the Internet Archive appeals the judgment (which it has indicated that it plans to do), and the appeals court overturns Judge Koeltl’s decision, for example by finding that its CDL program is a fair use, IA may be able to resume lending out those CDL versions of books published by the plaintiffs which are also available as ebooks. The agreement also does not mean that IA has to end its CDL program entirely—neither books published by other publishers nor books published by the publisher plaintiffs that are not available as ebooks are covered under the judge’s order.  

What’s Next?

The filing represents the first step towards the Internet Archive appealing the court’s judgment. As we’ve said before, Authors Alliance plans to write another amicus brief in support of the Internet Archive’s argument that Controlled Digital Lending is a fair use. Now that the judge has issued his final judgment, IA has 30 days to file a “notice of appeal” with the district court. Then, the case will receive a new docket in the Second Circuit Court of Appeals, and the various calendar and filing processes will begin anew under the rules of that court. We will of course keep our readers apprised of further developments in this case.

Ninth Circuit Issues Decision in Hunley v. Instagram

Posted July 19, 2023
Photo by Alexander Shatov on Unsplash

On Monday, the Ninth Circuit issued a decision in Hunley v. Instagram, a case about whether Instagram (and platforms like it) can be held liable for secondary infringement based on its embedding feature, whereby websites employ code to display an Instagram post on their sites within their own content. We are delighted to announce that the court ruled in favor of Instagram, reinforcing important copyright principles which allow authors and other creators to link to and embed third-party content, enriching their writing in the process. 

Our Brief

Authors Alliance signed on to an amicus brief in this case, arguing that Instagram should not be held liable for contributory infringement for its embedding feature. We explained that Instagram was not liable under a precedential legal test established in Perfect 10 v. Amazon, and moreover that a ruling to the contrary could place our ability to link to other online content (which is analogous to embedding in many ways) at risk for legal liability. 

Narrowing the Perfect 10 test—which establishes that a website does not infringe when it does not store a copy of the relevant work on its server—would have struck a blow to how we share and engage with online content. Linking to other information allows authors to easily cite to other information without disrupting the flow of their writing. By the same token, it allows internet users to verify information and learn more about topics of interest, all with the click of a button. We are pleased that the court ruled in favor of Instagram, declining to revisit the Perfect 10 test and holding that it foreclosed relief for the photographers that had filed the lawsuit. In so doing, the court has helped maintain a vibrant internet where all can share and engage with knowledge and creative expression.

The Decision

The case concerned a group of photographers whose instagram posts were embedded into content by several media outlets. The photographers then sued Instagram in the Northern District of California, on the theory that by offering the “embedding” feature, it was facilitating copyright infringement of others and therefore was liable. The district court found that Perfect 10 applied to the case, and therefore that Instagram was not liable for infringement for the outlets’ display of the posts. 

The Ninth Circuit agreed, and furthermore declined to revisit or narrow the Perfect 10 case for a number of reasons—it rejected the argument that the search engines at issue in the Perfect 10 case itself were somehow different from social media platforms, and affirmed that Perfect 10 was consistent with more recent Supreme Court case law. The court also cited with approval our argument that embedding and in-line linking have paved the way for innovation and creativity online, though did not adopt the justification, reasoning that it is not a court’s job to serve as a policymaker. In applying the Perfect 10 test, the court explained that Instagram did not infringe the photographers’ copyrights, and where there is no direct infringement, there cannot be related secondary infringement. Instagram displayed a copy of the relevant photographs on its platform, which users permit via a license they agree to by using the platform. But it did not facilitate the images’ display elsewhere, because the computer code used by the media platforms that embedded the instagram posts did not make a copy of the posts, but rather formatted and displayed them. 

Copyright Office Hosts Listening Session on Copyright in AI-Generated Audiovisual Works

Posted June 26, 2023
Photo by Jon Tyson on Unsplash

On May 17, the Copyright Office held a listening session on the topic of copyright issues in AI-generated audiovisual works. You may remember that we’ve covered the other listening sessions convened by the Office on visual arts, musical works, and textual works (in which we also participated). In today’s post, we’ll summarize and discuss the audiovisual works listening session and offer some insights on the conversation.

Participants in the audiovisual works listening session included AI developers in the audiovisual space such as Roblox and Hidden Door; trade groups and professional organizations including the Motion Picture Association, Writers Guild of America West, and National Association of Broadcasters; and individual filmmakers and game developers. 

Generative AI Tools in Films and Video Games

As was the case in the music listening session, multiple participants indicated that generative AI is already being used in film production. The representative from the Motion Picture Association (MPA) explained that “innovative studios” are already using generative AI in both the production and post-production processes. As with other creative industries, generative AI tools can support filmmakers by increasing the efficiency of various tasks that are part of the filmmaking process. For example, routine tasks like color correction and blurring or sharpening particular frames are made much simpler and quicker through the use of AI tools. Other participants discussed the ways in which generative AI can help with ideation, overcoming “creativity blocks,” eliminating some of the drudgery of filmmaking, enhancing visual effects, and lowering barriers to entry for would-be filmmakers without the resources of more established players. These examples are analogous to the various ways that generative AI can support authors, which Authors Alliance and others have discussed, like brainstorming, developing characters, and generating ideas for new works.

The representative from the MPA also emphasized the potential for AI tools to “enhance the viewer experience” by making visual effects more dramatic, and in the longer term, possibly enable much deeper experiences like having conversations with fictional characters from films. The representative from Hidden Door—a company that builds “online social role-playing games for groups of people to come together and tell stories together”—similarly spoke of new ways for audiences to engage with creators, such as by creating a sort of fan fiction world with the use of generative AI tools, with contributions from the author, the user, and the generative AI system. And in fact, this can create “new economic opportunities” for authors, who can monetize their content in new and innovative ways. 

Video games are similarly already incorporating generative AI. In fact, generative AI’s antecedents, such as “procedural content generation” and “rule-based systems” have been used in video games since their inception. 

Centering Human Creators

Throughout the listening session, participants emphasized the role of human filmmakers and game developers in creating works involving AI-generated elements, stating or implying that creators who use generative AI should own copyrights in the works they produce using these tools. The representative from Roblox, an online gaming platform that allows users to program games and play other users’ games, emphasized that AI-generated content is effective and engaging because of the human creativity inherent in “select[ing] the best output” and making other creative decisions. A representative from Inworld AI, a developer platform for AI characters, echoed this idea, explaining that these tools do not exist in isolation, but are productive only when a human uses them and makes creative choices about their use, akin to the use of a simpler tool like a camera or paintbrush. 

A concern expressed by several participants—including the Writers Guild of America West, National Broadcasters Association, and Directors Guild—is that works created using generative AI tools could devalue works created by humans without such tools. The idea of markets being “oversaturated” with competing audiovisual works raises the possibility that individual creators could be crowded out. While this is far from certain, it reflects increasing concerns over threats to creators’ economic livelihoods when AI-generated works compete alongside theirs. 

Training Data and Fair Use

On the question of whether using copyrighted training materials to train generative AI systems is a fair use, there was disagreement among participants. The representative from the Presentation Guild likened the use of copyrighted training data without permission to “entire works . . . being stolen outright.” They further said that fair use does not allow this type of use due to the commercial nature of the generative AI companies, the creative nature of the works used to train the systems (though it is worth noting that factual works, and others entitled only “thin” copyright protection, are also use to train these tools), and because by “wrest[ing] from the creator ownership and control of their own work[,]” the market value for those works is harmed . This is not, in my view, an accurate statement of how the market effects factor in fair use works, because unauthorized uses that are also fair always wrest some control from the author—this is part of copyright’s balance between an author’s rights and permitting onward fair uses. 

The representative from the Writers Guild of America (“WGA”) West—which is currently on strike over, among other things, the role of generative AI in television writing—had some interesting things to say about the use of copyrighted works as training data for generative AI systems. In contract negotiations, WGA had proposed a contract which “would . . . prohibit companies from using material written under the Guild’s agreement to train AI programs for the purpose of creating other derivative and potentially infringing works.” The companies refused to acquiesce, arguing that “the technology is new and they’re not inclined to limit their ability to use this new technology in the future.” The companies’ positions are somewhat similar to those expressed by us and many others—that while generative AI remains in its nascency, it is sensible to allow it to continue to develop before instituting new laws and regulations. But it does show the tension between this idea and creators who feel that their livelihoods may be threatened by generative AI’s potential to create works with less input from human authors. 

Other participants, such as the representative from Storyblock, a stock video licensing company, emphasized their belief that creators of the works used to train generative AI tools should be required to consent, and should receive compensation and credit for the use of their works to train these models. The so-called “three C’s” idea has gained traction in recent months. In my view, the use of training data is a fair use, making these requirements unnecessary from a copyright perspective, but it represents an increasingly prevailing view among rightsholders and licensing groups (including the WGA, motivating its ongoing strike in some respects) when it comes to making the use of generative AI tools more ethical. 

Adequacy of Registration Guidance

Several participants expressed concerns about the Copyright Office’s recent registration guidance regarding works containing AI-generated materials, and specifically how to draw the line between human-authored and AI-generated works when generative AI tools are used as part of a human’s creative process. The MPA representative explained that the guidance does not account for the subtle ways in which generative AI tools are used as part of the filmmaking process, where it often works as a component of various editing and production tools. The MPA representative argued that using these kinds of tools shouldn’t make parts of films unprotected by copyright or trigger a need to disclose minor uses of such tools in copyright registration applications. The representative from Roblox echoed these concerns, noting that when a video game involves thousands of lines of code, it would be difficult for a developer to disclaim copyright in certain lines of code that were AI-generated. 

A game developer and professor expressed her view that in the realm of video games, we are approaching a reality where generative AI is “so integrated into a lot of consumer-grade tools that people are going to find it impossible to be able to disclose AI usage.” If creators or users do not even realize they are using generative AI when they use various creative digital tools, the Office’s requirement that registrant’s disclose their use of generative AI in copyright registration applications will be difficult if not impossible to follow.

Supreme Court Announces Decision in Jack Daniel’s v. VIP Products

Posted June 8, 2023
Photo by Chris Jarvis on Unsplash

Today, the Supreme Court handed down a decision in Jack Daniel’s v. VIP Products, a trademark case about the right to parody popular brands, for which Authors Alliance submitted an amicus brief, supported by the Harvard Cyberlaw clinic. In a unanimous decision, the Court vacated and remanded the Ninth Circuit’s decision, overturning the decision asking the lower courts to re-hear the case with a new, albeit it very narrow, principle announced by the Court: special First Amendment review is not appropriate in cases where one brand’s trademark is used in another, even when used as a parody. In addition to the majority opinion delivered by Justice Kagan, there were two concurring opinions by Justice Sotomayor and Justice Gorsuch, each joined by other justices. 

Case Background

The case concerns a dog toy that parodies Jack Daniel’s famous Tennessee Whiskey bottle, using some of the familiar features from the bottle, and bearing the label “Bad Spaniels.” After discovering the dog toy, Jack Daniel’s requested that VIP cease selling the toys. VIP Products refused, then proceeded to file a trademark suit, asking for a declaratory judgment that its toy “neither infringed nor diluted Jack Daniel’s trademarks.” Jack Daniel’s then countersued to enforce its trademark, arguing that the Bad Spaniels toy infringed its trademark and diluted its brand. We became interested in the case because of its implications for creators of all sorts (beyond companies making humorous parody products). 

As we explain in our amicus brief, authors rely on their ability to use popular brands in their works. For example, fiction authors might send their characters to real-life colleges and universities, set scenes where characters dine at real-life restaurant chains, and use other cultural touchstones to enrich their works and ultimately, to express themselves. While the case is about trademark, the First Amendment looms large in the background. A creator’s right to parody brands, stories, and other cultural objects is an important part of our First Amendment rights, and is particularly important for authors. 

Trademark law is about protecting consumers from being confused as to the source of the goods and services they purchase. But it is important that trademark law be enforced consistent with the First Amendment and its guarantees of free expression. And importantly, trademark litigation is hugely expensive, often involving costly consumer surveys and multiple rounds of hearings and appeals. We are concerned that even the threat of litigation could create a chilling effect on authors, who might sensibly decide not to use popular brands in their works based on the possibility of being sued. 

In our brief, we suggested that the Court implement a framework like the one established by the Second Circuit in Rogers v. Grimaldi, “a threshold test . . . designed to protect First Amendment interests in the trademark context.” Under Rogers, in cases of creative expressive works, trademark infringement should only come into play “where the public interest in avoiding consumer confusion outweighs the public interest in free expression.” It establishes that trademark law should only be applied where the use “has no artistic relevance to the underlying work whatsoever, or, if it has some artistic relevance, unless the [second work] explicitly misleads as to the source or the content of the work.”

The Supreme Court’s Decision

In today’s decision, the Court held that “[w]hen an alleged infringer uses a trademark as a designation of source for the infringer’s own goods, the Rogers test does not apply.” Without directly taking a position on the viability of the Rogers test, the Court found that, in this circumstance, where it believed that VIP Products used Jack Daniel’s trademarks for “source identifiers,” the test was inapplicable. It held that the Rogers test is not appropriate when the accused infringer has used a trademark to designate the source of its own goods—in other words, has used a trademark as a trademark.” The fact that the dog toy had “expressive content” did not disturb this conclusion. 

Describing Rogers as a noncommercial exclusion, the Court said that VIP’s use was commercial, as it was on a dog toy available for sale (i.e. a commercial product). Further supporting this conclusion, the Court pointed to the fact that VIP Products had registered a trademark in “Bad Spaniels.” It found that the Ninth Circuit’s interpretation of the “noncommercial use exception” was overly broad, noting that the Rogers case itself concerned a film, an expressive work entitled to the highest First Amendment protection, and vacating the lower court’s decision. 

The Court instead directed the lower court to consider a different inquiry, whether consumers will be confused as to whether Bad Spaniels is associated with Jack Daniel’s, rather than focusing on the expressive elements of the Bad Spaniels toy. But the Court also explained that “a trademark’s expressive message—particularly a parodic one, as VIP asserts—may properly figure in assessing the likelihood of confusion.” In other words, the fact that the Bad Spaniels toy is (at least in our view) a clear parody of Jack Daniel’s may make it more likely that consumers are not confused into thinking that Jack Daniel’s is associated with the toy. In her concurrence, Justice Sotomayor underscored this point by cautioning lower courts against relying too heavily on survey evidence when deciding whether consumers are confused “in the context of parodies and potentially other uses implicating First Amendment concerns.” In so doing, Justice Sotomayor emphasized the importance of parody as a form of First Amendment expression. 

The Court’s decision is quite narrow. It does not disapprove of the Rogers test in other contexts, such as when a trademark is used in an expressive work, and as such, it is unlikely to have a large impact on authors using brands and marks in their books and other creative expression. Lower courts across the country that do use the Rogers test may continue to do so under VIP Products.However, Justice Gorsuch’s concurrence does express some skepticism about the Rogers test and its applications, cautioning lower courts to handle the test with care. However, as a concurrence, this opinion has much less precedential effect than the majority’s. 

Remaining Questions

All of this being said, the Court does not explain why First Amendment scrutiny should not apply in this case, but merely reiterates that Rogers as a doctrine is and has always been “cabined,” with  “the Rogers test [applying] only to cases involving ‘non-trademark uses[.]’” The Court relies on that history and precedent rather than explaining the reasoning. Nor does the Court discuss the relevance of the commercial/noncommercial use distinction when it comes to the role of the First Amendment in trademark law. In our view, the Bad Spaniels toy did contain some highly expressive elements and functioned as a parody, so this omission is significant. And it may create some confusion for closer cases—at one point, Justice Kagan explains that “the likelihood-of-confusion inquiry does enough work to account for the interest in free expression,” “except, potentially, in rare situations.” We are left to wonder what those are. She further notes that the Court’s narrow decision “does not decide how far the ‘noncommercial use exclusion’ goes.” This may leave lower courts without sufficient guidance as to the reach of the noncommercial use exclusion from trademark liability and what “rare situations” merit application of the Rogers test to commercial or quasi-commercial uses.

Copyright Office Holds Listening Session on Copyright Issues in AI-Generated Music and Sound Recordings

Posted June 2, 2023
Photo by Possessed Photography on Unsplash

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. 

Supreme Court Issues Decisions in Warhol Foundation and Gonzalez

Posted May 19, 2023
Photo by Claire Anderson on Unsplash

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. 

Authors Alliance Welcomes Christian Howard-Sukhil as Text Data Mining Legal Fellow

Posted May 12, 2023

As we mentioned in our blog post on our Text Data Mining: Demonstrating Fair Use project a few weeks back, Authors Alliance is pleased to have Christian Howard-Sukhil on board as our brand new Text Data Mining legal fellow. As part of our project, generously funded by the Andrew W. Mellon Foundation, we established this new fellowship to provide research and writing support for our project. Christian will help us produce guidance for researchers and a report on the usability, successes, and challenges of the text data mining exemption to Section 1201’s prohibition on bypassing technical protection measures that Authors Alliance obtained in 2021. Christian begins her work with Authors Alliance this week, and we are thrilled to have her. 

Christian holds a PhD in English Language and Literature from the University of Virginia, and has just completed her second year of law school at UC Berkeley. Christian has extensive digital humanities and text data mining experience, including in previous roles at UVA and Bucknell University. Her work with Authors Alliance will focus on researching and writing about the ways that current law helps or hinders text and data mining researchers in the real world. She will also contribute to our blog—look out for posts from her later this year.

About her new role at Authors Alliance, Christian says, “I am delighted to join Authors Alliance and to help support text and data mining researchers navigate the various legal hurdles that they face. As a former academic and TDM researcher myself, I saw first-hand how our complicated legal structure can deter valid and generative forms of TDM research. In fact, these legal issues are, in part, what inspired me to attend law school. So being able to work with Authors Alliance on such an important project—and one so closely tied to my own background and interests—is as exciting as it is rewarding.”

Please join us in welcoming Christian!

Copyright Office Holds Listening Session on Copyright Issues in AI-Generated Visual Works

Photo by Debby Hudson on Unsplash

Earlier this week, the Copyright Office convened a second listening session on the topic of copyright issues in AI-generated expressive works, a part of its initiative to study and understand the issue, and following its listening session on copyright issues in AIgenerated textual works a few weeks back (in which Authors Alliance participated). Tuesday’s sessions covered copyright issues in images created by generative AI programs, a topic that has garnered substantial public attention and controversy in recent months.

Participants in the listening sessions included a variety of professional artist organizations like National Press Photographers Association, Graphic Artists Guild, and Professional Photographers of America; companies that have created the generative AI tools under discussion, like Stability AI, Jasper AI, and Adobe; several individual artists; and a variety of law school professors, attorneys, and think tanks representing varied and diverse views on copyright issues in AI-generated images. 

Generative AI as a Powerful Artistic Tool

Most if not all of the listening sessions’ participants agreed that generative AI programs had the potential to be incredible tools for artists. Like earlier technological developments such as manual cameras and, much more recently, image editing software like Photoshop, generative AI programs can minimize or eliminate some of the “mechanical” aspects of creation, making creation less time-consuming. But participants disagreed on the impact these tools are having on artists and whether the tools themselves or copyright law ought to be reformed to address these effects. 

Visual artists, and those representing them, tended to caution that these tools should be developed in a way that does not hurt the livelihoods of the artists who created the images the programs are trained on. While a more streamlined creative process makes things easier for artists relying on generative AI in their creation, it could also mean fewer opportunities for others artists. When a single designer can easily create background art with Midjourney, for example, they might not need to hire another designer for that task. This helps the first designer to the detriment of the second. Those representing the companies that create and market generative AI programs, including Jasper AI and Stability AI, focused on the ways that their tools are already helping artists: these tools can generate inspiration images as “jumping off points” for visual artists and lower barriers to entry for aspiring visual artists who may not have the technical skills to create visual art without support from these kinds of tools, for example. 

On the other hand, some participants voiced concerns about ethical issues in AI-generated works. A representative from the National Press Photographers Association mentioned concerns that AI-generated images could be used for “bad uses,” and creators of the training data could be associated with these kinds of uses. Deepfakes and “images used to promote social unrest” are some of the uses that photojournalists and other creators are concerned about. 

Copyright Registration in AI-Generated Visual Art

Several participants expressed approval of the Copyright Office’s recent guidance regarding registration in AI-generated works, but others called for greater clarity in the registration guidance. The guidance reiterates that there is no copyright protection in works created by generative AI programs, because of copyright’s human authorship requirement. It instructs creators that they can only obtain copyright registration for the portions of the work they actually created, and must disclose the role of generative AI tools in creating their works if it is more than de minimis. An author can also obtain copyright protection for a selection and arrangement of AI-generated works as a compilation, but not in the AI-generated images themselves. Yet open questions, particularly in the context of AI-generated visual art, remain: how much does an artist need to add to an image to render it their own creation, rather than the product of a generative AI tool? In other words, how much human creativity is needed to transform an AI-generated image into the product of original human creation for the purposes of copyright? How are we to address situations where a human and AI program “collaborate” on the creation of a work? The fact that the Office’s guidance requires applicants to disclose if they used AI programs in the creation of their work also leaves open questions. If an artist uses a generative AI program to create just one element of a larger work, or as a tool for inspiration, must that be disclosed in copyright registration applications? 

The attorney for Kristina Kashtanova, the artist who applied for a copyright registration for her graphic novel, Zarya of the Dawn also spoke. If you haven’t been tracking it, Zarya of the Dawn included many AI-generated images and sparked many of the conversations around copyright in AI-generated visual works (you can read our previous coverage of the Office’s decision letter on Zarya of the Dawn here). Kashtanova’s attorney raised more questions about the registration guidance. She pointed out that the amount of creativity required to create a copyrighted work is very low—there must be more than a “modicum” of creativity, meaning that vast quantities of works (like each of the photographs we take with our smartphones) are eligible for copyright protection. Why, then, is the bar higher when it comes to AI-generated works? Kashantova certainly had to be quite creative to put together her graphic novel, and the act of writing a prompt for the image generator, refining that prompt, and re-prompting the tool until the creator gets an image they are satisfied with requires a fair amount of creative human input. More, one might argue, than is required to take a quick digital photograph. The registration guidance attempts to solve the problem of copyright protection in works not created by a human, but in so doing, it creates different copyrightability standards for different types of creative processes. 

These questions will become all the more relevant as artists increasingly rely on AI programs to create their works. The representative from Getty Images stated that more than half of their consumers now use generative AI programs to create images as part of their workflows, and several of the professional artist organizations noted that many of their members were similarly taking up generative AI tools in their creation.

Calls For Greater Transparency

Many participants expressed a desire for the companies designing and making available generative AI programs to be more transparent about the contents of these tools’ training data. This appealed both to artists who were concerned that their works were used to train the models, and felt this was fundamentally unfair, and those with ethical concerns around scraping or potential copyright infringement. Responsive to these critiques, Adobe explained that it sought to develop its new AI image generator, Firefly (which is currently in beta testing) in a way that is responsive to these kinds of concerns. Adobe explained that it planned to train its tool on openly licensed images, seeking to “drive transparency standards” and “deploy [the] technology responsibly in a way that respects creators and our communities at large.” The representative from Getty Images also called for greater transparency in training data. Getty stated that transparency could help mitigate the legal and economic risks associated with the use of generative AI programs—potential copyright claims as well as the possibility of harming the visual artists who created the underlying works they are trained on. 

Opt-Outs and Licensing 

Related to calls for transparency, much of the discussion centered around attempts to permit artists to opt out of having their works included in the training data used for generative AI programs. Like robots.txt, a tag that allows websites to indicate to web crawlers and other web robots that they don’t wish to allow these robots to visit their sites, several participants discussed a “do not train tag” as a way for creators to opt out of being included in the training data. Adobe said it intended to train its new generative AI tool, Firefly, on openly licensed images and make it easy for artists to opt out with a “do not train” tag, apparently in response to these types of concerns. Yet some rightsholder groups pointed out that compliance with this tag may be uneven—indeed, robots.txt itself is a voluntary standard, and so-called bad robots like spam bots often ignore it. 

Works available under permissive licenses like Creative Commons’ various licenses have been suggested as good candidates for training data to avoid potential rights issues. Though several participants pointed out that there may be compliance issues when it comes to commercial uses of these tools, as well as attribution requirements. And the participant representing the American Society for Collective Rights Licensing voiced support for proposals to implement a collective licensing scheme to compensate artists whose works are used to train generative AI programs, echoing earlier suggestions by groups such as the Authors Guild. 

One visual artist argued fervently that an opt out standard was not enough: in her view, visual artists should have to opt in to having their works included in training data, as, in her view, an opt out system harms artists without much of an online presence or the digital literacy to affirmatively opt out. In general, the artist participants voiced strong opposition to having their works included without compensation, a position many creators with concerns about generative AI have taken. But Jasper AI expressed its view that training generative AI programs with visual works found across the Internet was a transformative use of that data, all but implying that this kind of training was fair use (a position Authors Alliance has taken). It was notable that so few participants suggested that the ingestion of visual works of art for the purposes of training generative AI programs was a fair use, particularly compared to the arguments in the listening session on text-based works. This may well be due to ongoing lawsuits, inherent differences between image based and text based outputs, or the general tenor of conversations around AI-generated visual art. Many of the participants spoke of anecdotal evidence that graphic artists are already facing job loss and economic hardship as a result of the emergence of AI-generated visual art.