Jim Neal

Welcome to the Library Journal Roundtable. The theme for today is copyright. The context is libraries. My name is Jim Neal. I’m University Librarian Emeritus at Columbia University in New York and Senior Policy Fellow at the American Library Association. I will serve as the moderator.

Allow me to introduce the members of the panel. Jonathan Band is the counsel to the Library Copyright Alliance. He works with the American Library Association and the Association of Research Libraries. Sara Benson is Associate Professor and Copyright Librarian at the University of Illinois Library. She’s also an affiliate professor at the School of Information of the Siebel Center for Design, the European Union Center and the Center for Global Studies. Rick Anderson is the University Librarian at Brigham Young University. Kyle Courtney is Director of Copyright and Information Policy at Harvard and founder of two library nonprofits, Library Futures and the eBook Study Group.

All of these individuals are copyright and information policy experts with years and years of deep involvement in education and advocacy around the importance of copyright for libraries, the laws and legislation which influence our work in libraries.  

Jonathan, I’d like you to lead us off and tell us a little bit how you look at the past year in terms of copyright.

Jonathan Band

Thank you very much for that introduction, Jim. Perhaps the overarching issue in my view over the past year has been all of the litigation relating to artificial intelligence. At this point, more than 70 cases have been filed in the United States. And interestingly, we’re starting to see some decisions. I’ll just confine my remarks to that one case, and this is Bartz v. Anthropic. The main reason we’ve seen decisions in that case is that the judge, Judge William Alsup in the Northern District of California, had announced that he was retiring at the end of 2025. He decided that he was going to retire, and he was going to have things wrapped up nicely by the time he retired, and he did. So the case proceeded at a much more accelerated schedule than it otherwise would. And there are three interesting and important aspects of the way he decided the case.

First, on basically on motions for summary judgment, he decided that the scraping of internet sites for content that Anthropic did in the course of training its a large language model, its LLM, was highly transformative and ultimately concluded that that was a fair use. Then interestingly, just a few days later, another judge in the Northern District in essence agreed with him on that conclusion in a separate case, the Kadrey case.

I think that there’s a good chance that other courts will similarly follow Judge Alsup’s ruling. It was pretty well reasoned, even though, as I said, it was kind of hurried. He’s a relatively well-respected judge, so I think that there's a good chance that other courts are going to be influenced by his ruling on that. But again, 70 cases, a lot of them are going to be looking at this issue about whether the copying that occurs in the course of training, whether that is a fair use or not, and so it’s certainly possible that other courts will rule differently, and of course on appeal, who knows what will happen.

The second ruling was that even though the general scraping of websites for training purposes was a fair use, the copying of “shadow libraries,” various libraries or online websites set up for infringing purposes, that their harvesting of materials from those kinds of websites was not a fair use, and it’s unclear again, because of the hurried nature of the decision, he seems to have concluded that Anthropic was not harvesting that material simply for training purposes, but was actually planning on keeping it for other purposes.

It would not surprise me if other courts reached that similar conclusion that scraping of publicly available materials on a website, material that was placed with authorization by the copyright owners—that’s okay. On the other hand, getting material from a pirate site, that’s not okay; to some extent, that's sort of like a Solomonic ruling that judges like, and especially if you’re viewing fair use as a equitable rule of reason, one could sort of see that that’s a logical outcome. Interestingly, it has turned out that many of the AI firms have in fact trained on pirate sites. That’s something that was not widely understood previously, but now seems to be the case, and so that could be an important fact in future litigation.

Where I think Judge Alsup sort of went off the rails is that the case was brought as a class action, so there were three authors bringing their case relating to their specific works, but they brought the case as a class action. Judge Alsup agreed to certify the class. He agreed that it should proceed basically, that they would be representing not just themselves, those three authors, but all authors whose works were scraped by Anthropic. And that decision, making it a class action instead of just an individual action, sort of exploded the case. Even though the court ruled that the basic scraping of these open websites was fair use, the nber of works at issue on these pirate sites is still enormous.

The exposure that Anthropic was facing by virtue of it now being a certified class as opposed to an individual action, were enormous in the trillions of dollars, literally trillions with a T. And so Anthropic had little choice but to settle the case. There’s been a lot of back and forth since they settled the case that, in my view, makes certifying a class a mistake. Once Judge Alsup certified the class, I really think Anthropic had no choice but to settle.

All the reasons why he shouldn’t have certified the case are becoming abundantly clear in all the wrangling and, and argents between the various attorneys representing the various plaintiffs, with each other, with the court, with the defendant—it’s all becoming apparent. The problem is that determining corporate ownership is a complicated business. And especially with books, you’re dealing with older books, books where the publishers have gone out of business, where the authors are dead, where there are heirs, different kinds of books—academic books, educational books, trade books—there are so many issues. There’s going to be many trials. It's going to be extremely complicated. This was pointed out to judge, but he sort of ignored it, and I think that was a mistake.

Thank you, Jonathan. Sara, what issue would you identify from the past year that we should be focused on? 

Sara Benson

Thank you so much for the introduction and the invitation to join this esteemed panel. I would point out that I agree with Jonathan that artificial intelligence and generative artificial intelligence is kind of the issue of the day in terms of copyright. It’s the number one inquiry that I see on a day-to-day basis as a copyright librarian.

In terms of how I view this issue in research libraries, and that’s the place where I sit professionally, I think the most impactful thing from the past year from our perspective is the U.S. Copyright Office’s recent three-part advisory guidelines on copyright and AI—in particular their most recent guidance on generative AI training. It’s still labeled a pre-publication and there's a lot swirling around about it, especially since the head of the Copyright Office was fired and, according to her, largely in response to the contents of that report.

This report is really important for researchers and those who are doing nonprofit research with scraping different copyrighted materials. In the report the Copyright Office takes the position—and this really isn't a surprise—that the more the work that you’re doing is research-related and non-profit-related, the more likely it is to be a fair use, especially when you are doing something as transformative as generating new purpose, meaning, or message with generative AI. And, so, I think for the folks that I work with who are really trying to do innovative and new things with generative AI, I think this report is quite important because researchers and research institutions tend to be pretty risk-averse when it comes to scraping and copyright. The report was really important because it gave them a lot of power in terms of going forward the types of things that they can do with a little less fear and with the support of that report.

I think that’s really crucial because we really do need to keep up with the big tech companies. They are also scraping work, but for commercial purposes. And we have people doing research for medical purposes, for different things that are really aiming to improve the world. We need to also feel that we have that power as researchers, as research institutions, to make a positive change using generative AI.

I know in the library community there’s also a lot of people have some problems with AI in terms of ethical concerns, plagiarism concerns, environmental concerns. I think all these things are valid but I also think it’s important that we recognize that there are some really innovative things that are coming out of AI and that we as colleges, universities, librarians who sit in these places, we need to also feel empowered to help our researchers make those important discoveries and not to say that copyright law is standing in their way. I encourage people to go to the U.S. Copyright Office and to read that report because I think it’s a positive impact for the research community, and hopefully librarians can empower the researchers from where they sit in the library as well.  

Kyle, can you continue with your thoughts? 

Kyle Courtney

Absolutely. So, you know, we had a plethora of things to choose from this time. And I chose the question about whether fair use or other core copyright exceptions that are relied on frequently by libraries and archives and our colleagues and cultural institutions can survive in a world governed by licenses. This is about copyright’s interaction with the license, where the short answer that I’m seeing is that licensing is increasingly being treated as sort of a substitute for those copyright rights. In practice, it’s overriding the very exceptions that are supposed to protect libraries and their users.

A clear example would be the Internet Archive ebook litigation. In that case—and I was involved in that case—the courts didn’t just reject control of digital lending, but they relied heavily on an idea that publishers already offer ebooks through licensing markets, and therefore those licensing markets should control access. This is despite the fact that Internet Archive and every other library in the country lawfully acquired physical books that they want to do certain things with under those various exceptions, including interlibrary loan, preservation copies, including making fair uses of it.

The logic effectively elevated the fourth factor, market harm, above all the others, I thought, by defining harm in relation to licensed markets that the publishers themselves created. So, the result is troubling. If a rights holder chooses to distribute works through restrictive licenses, courts may treat that licensing market as determinative, even when the library is a nonprofit, even when access is limited, even when the use aligns with traditional library functions. In other words, the existence of a license becomes the reason for fair use for sale. Even preservation rights are treated as if they no longer apply.

This dynamic has broader implications. It doesn’t just affect fair use. I think it undermines Section 108 and 109. There’s no digital first sale, full stop. Libraries can’t rely on statutory rights if they never truly own digital works in the first place, right? So copyright’s public interest safeguards become theoretical in that view, when licensing terms become the real law on the ground.

So, what should we be doing? You know, maybe we should stop treating licenses as neutral or inevitable. Licensing, I believe, is a policy choice. And when it waives statutory rights, it should be challenged in negotiations and institutional policies and in public discourse, which we’re doing here today. 

I think we need to reframe the conversation on what happened this year in the past. I want to be clear, this isn’t about where libraries want special treatment. You know, we are special, obviously, but we don’t ask you for special treatment. It’s about whether copyright law as enacted by Congress still matters in the digital environment, or whether it’s just quietly been replaced by private contracts that erase the balance that exists in copyright law that was designed to maintain and maintains the library mission. So that would be my take on the past, and looking forward to other folks.

Rick, what is the development of the past year that you would highlight?

Rick Anderson

I’m going to zoom in on one particular thing that happened this last year that I think, in many ways, we might see as a very obvious development, but that is important because it created new case law—that is the decision in the Thaler v. Perlmutter case. This was a case that established that an AI bot or an AI agent can't hold copyright because only human beings are recognized under the law as authors and creators.

In 2024, a computer scientist named Steven Thaler invented a generative AI bot that he called the “Creativity Machine,” and he used it to create a picture that he called, “A recent entry to paradise.” Then he applied for copyright registration for the picture. In his application form, he indicated that the Creativity Machine was the work’s creator and that he himself was the work’s owner. And the U.S. Copyright Office denied the copyright registration request because copyright law “only protects the fruits of intellectual labor that are founded in the creative powers of the mind and that reflect original intellectual conceptions of the author.” So Thaler sued.

He acknowledged that the work had no traditional human author, which I thought was a cute phrase. But he argued that the human authorship requirement in current copyright law is out of date in light of all the recent technological developments. He also argued that the work should be created as a work for hire. So, in other words, he should be considered the copyright holder in this AI-generated work the same way he would be if he had paid someone else to create it under a work for hire agreement. And he also argued on appeal that he should be the copyright holder because he had both invented the Creativity Machine and had fashioned the prompt that the Creativity Machine used to create the image. Now, the original district court rejected his claims saying that “human authorship is a bedrock requirement of copyright,” and also rejected the work for hire argent because the Creativity Machine had not created a copyrightable work. Thaler couldn’t therefore argue that copyright had been transferred to him from the creator because there was no copyrightable work.

He appealed the district court’s decision, but it was upheld by the appellate court. The court of appeals affirmed that “authors are at the center of the copyright act,” and pointed out that, “numerous copyright act provisions both identify authors as human beings and define machines as tools used by human beings in the creative process rather than as creators themselves.”

The appeals court laid out what struck me as a very interesting and cogent list of reasons why it doesn't make sense to consider a machine an author, including the fact that, well, all of them basically boil down to the fact that a machine just is not a person. Therefore, we can’t have what the law considers essential characteristics of authorship. For example, a machine cannot hold property. It’s hard to be a copyright holder if you’re incapable of holding property.  A machine can't have intentions and doesn’t have legal rights and therefore can’t transfer its rights in something to someone else. Existing copyright law consistently characterizes machines as tools rather than as agents. So, I think that the Thaler v. Perlmutter case is incredibly important and also leaves unexamined a bunch of questions that are going to just proliferate in the coming years regarding the degree to which AI can contribute to copyrightable creations.

Thank you, Rick. I think we should spend a few minutes digging a bit deeper into some of these topics.

Kyle, you talked about the expansion of licensing undermining copyright, overriding the library exceptions in particular, and the user rights. Could you talk a little bit about what the library profession should be doing to address and deal with some of these developments?

Kyle Courtney 

Yeah, libraries need to act now because I think their mission is being delineated to them by another party. We see this in the ebooks. This is funny because we’re talking about copyright, but you know, you can’t talk about copyright without talking about licensing and contract law. Libraries could face archival loss, inability to lend, restricted research use, or growing dependence on proprietary platforms that were never designed for something that we were designed for, long-term stewardship and public access. If we can imagine a cultural memory problem or a democratic access problem, this is it.

In my mind, we have something that also happened this year that was not a copyright case that might be pending. This is some of my work with the Ebook Study Group against the backdrop of all this licensing. An important development in 2025 didn’t come from the courts though. It came from state legislatures. In May of 2025, Connecticut passed landmark ebook legislation, the first law of its kind, aimed squarely at addressing the restrictions that appear in these licensed ebook contracts.

The Connecticut law said, ‘Well, you’re doing well in this state, publishers. You're making lots of money. The libraries will continue to pay you for all this stuff, but the elimination of our mission through these contractual clauses should not stand.’ And so the law in Connecticut seeks to ensure that these library ebook licenses are aligned with the core library mission. And that’s a simple thing to do, but it’s hard for all the parties to come to a consensus here, because somehow there was a shift in the past and now it’s haunting us in the future that the libraries are competitive with the publishers. Somehow we can’t coexist when we have for a long time.

So, when markets and contracts fail to respect the public interests, the legislators can step in to restore the balance. The libraries in Connecticut came forth, and they delivered. If we can support legislative and judicial efforts that reaffirm the copyright exceptions—that they cannot be contracted away.

We see several other states considering this because they also want to reframe the conversation. Again, the conversation is not libraries want special treatment. It’s about libraries still want to exist in a digital environment and these licenses are causing us to prevent our services from happening. I think we need to push back against private contracts that erase that balance. I think for now, state action appears to be something that is moving forward. But the ultimate goal, to be honest, is if enough states pass this law, then maybe the federal government should come in and maybe do something.

Of course, I’ve had discussions with everyone on this call about do we amend Section 108? Do we amend Section 109? How do we get the library mission respected? These are big, giant conversations.  But I think what we can do now, what library professionals can do now, is support the movement for preservation and access to be more reflected in these licenses that libraries are dealing with.

We're seeing some interesting models developed in the global library community as well. Jonathan, why should library professionals care about AI and its implications? And is copyright the wrong tool for regulating AI? 

Jonathan Band

Thanks for that question. So, AI is everywhere. It is transcribing our conversation here as we speak. It is already permeating every aspect of our work, and that is the case with librarians.

AI, or certainly generative AI, is an amazing research tool. It is going to be a tool that students and faculty and people at public libraries, everyone, is going to be using this tool. And librarians are going to be helping them use the tool, instructing them around questions like what is it good for, what is it not good for, what are the shortcomings? AI is a research tool, a means of getting information, but it is getting information in kind of a strange way. It's probabilistic information. It’s not real information. And so that’s important for people to understand.

It’s also going to be the kind of thing that library users are going to be routinely using, asking AI to summarize materials that might be borrowed from the library. It’s going to be part of librarians’ workflow, whether it’s going to be in purchasing decisions or personnel decisions or just conducting business. So, it’s important to understand this technology, the copyright issues that are related to it, and its strengths and weaknesses. 

When I was first talking about litigation, there’s the question of copyright issues in the creation of AI. Then there’s also this sort of existential angst about whether AI is going to replace us and all of our jobs. And it might do that, or it might not replace all of our jobs, but it might change all of our jobs.

A lot of people are latching on to copyright as a way of sort of channeling AI, restricting or limiting it. But I think that that’s a mistake because copyright has very specific purpose of providing an incentive for authors to create works. It’s certainly not clear the impact that AI will have on that incentive.

It’s a tool that could be used, but it’s not a tool that’s designed necessarily for the purposes people want to use it for. It’s important to talk about how we want to regulate AI, but copyright and its defaults and its approaches are not the right tool. It’s too blunt a tool for coming up with the best solutions to the possible adverse impacts of AI.  

Rick, in light of the Thaler ruling, which you discussed, how do you think AI might affect future copyright rulings that deal particularly with more complex human-bot types of productions? 

Rick Anderson

As I mentioned before, the question that was settled in Thaler is really important, but in a way, I think it’s deceptively simple. The much more difficult question going forward is going to be how much human input into the creation of an original work is necessary in order for that work to be subject to copyright.

To illustrate how complicated that question might be, let me use an example from my own life. One of my hobbies is music production. And in music production, I use a variety of AI-based electronic tools. If I write an original song, and I create an original arrangement of that song, and I perform and record the instrumental parts, and then I use an AI-based tool to help me shape the sound of those recordings, I don’t think there’s any question that what I’ve created is still an original copyrightable work. The music is my creation, the arrangement is my creation, and the recordings of the performances are my creation. If I use AI during the mixing and the production process, that doesn’t somehow poison the copyrightability of my work because obviously the work is still original, and it’s the intentional work of a human author.  

However, interestingly, there are also now AI-based tools available that will create music according to prompts that I provide. If I’m writing a song and recording it, I can get an AI tool and tell it that I need a drum track that’s 120 bars long. It needs to be structured on a one-drop reggae pattern at a tempo of 80 beats per minute, it should have a modest number of drum fills, and it should have a two-beat kickoff prior to the first bar. Now, if an AI machine generates that drum track for me, it seems clear that I can’t assert copyright. That doesn’t mean that I can’t copyright the song—the song itself is still my original work—but what I think it does mean is that if somebody else strips that drum track out of my recording, they can then use that drum track freely without my permission because the drum track itself isn’t subject to copyright.  

Even here we’re in a fairly straightforward legal situation, but what if I also use an AI tool to generate the guitar part and an AI tool to generate the bass part and the keyboard parts? Again, I’m giving the AI machine some basic parameters to work with, but I’m not creating the melodies and the chord progressions myself. And then for that matter, what if I use AI to write the lyrics for me? Thaler would seem to indicate that the resulting song would not be subject to copyright, but what if I use all of those tools and then I take those outputs and I adapt them and change them and edit them?

How much adaptation on my part as a human being would it take for this AI-generated song to turn into something that the law would recognize as an original work? It kind of seems like as long as the final product is genuinely a variant on the original AI production, then maybe what I’ve created is a derivative work of a non-copyrightable original work, and therefore maybe not copyrightable, but at some point questions like these are going to make it to court, and I don’t think the court’s decisions are going to be slam dunks like Thaler v. Perlmutter, and that’s all just kind of a long-winded way of saying that I think we’re headed for some very interesting and very difficult times in terms of trying to figure out how AI and copyright can interact under the law.

Sara, we find the Copyright Office in the Library of Congress. And, as you’ve written about, there’s lots of interesting issues around it, relocating elsewhere in federal structure. How do you think the Copyright Office, regardless of where it may end up, might respond to any future legislative proposals around regulating AI? What’s the role of the Copyright Office looking forward in this area? 

Sara Benson

I think that Rick has laid out where the Copyright Office feels comfortable in the area of AI in terms of making real decisions—and that is where the Copyright Office has always been comfortable, which is copyright registration. They are the premier deciders of whether something is registrable or not; whether it’s sufficiently original or not. They have a very large compendium of their office practices, and they’ve put out guidance, and they’ve even gone to court recently about those issues.

Where I think that they are going to steer clear is in these decisions about fair use or whether they should give guidance to Congress or recommend that Congress pass laws governing AI. We have seen that this is happening around the world. In the EU, they have a copyright directive about artificial intelligence. U.S. companies that are doing business in the EU have to comply with all of these regulations, and they’re very specific.

But the United States, to date, has avoided entering into the law a specific legal regime about the copyright in AI in terms of what is a fair use. And I think that that’s for good reason: fair use is a flexible test. It is a case-by-case test, and it is one where the courts rule. As Jonathan pointed out, there are about 70 cases pending in the United States. We are in the throes of the decision making process, which isn’t usually a very fast process. It takes often years to make it through to a final decision, and a lot of cases will settle along the way. So we will not necessarily have 70 decisions. We will see a lot of those go away through settlement, likely. But we are starting to see some guidance from the courts about whether they think something is a fair use or not in terms of scraping. And then we’ll see some decisions on infringement as well on the output. And I think that that is where things should be headed. We should be watching the courts. We should be learning from the courts. And I think the Copyright Office also agrees with that. They think that this is a place where fair use is well suited to make that correct balance between whether something is illegal or copyright infringement, or it’s a fair use of a work.

I think that that makes the United States actually well suited to move ahead in the AI framework as opposed to other countries that are trying to regulate that really can slow progress quite a bit. I think we’re seeing an explosion of AI in this country in part because we have fair use, and we have the ability for courts to figure out what is the correct balance here instead of relying on Congress to lay out black and white rules. I think that would be a mistake, and I think that the Copyright Office agrees with that position.

Thank you, Sara. Let’s conclude with copyright issues and developments looking forward looking forward to 2026 and beyond. Let’s start with Rick.

Rick Anderson

So I’m hopefully not going to be too controversial, but when I think about looking into the future, specifically about the role of libraries with regard to copyright, I feel like the time has really come for us as a profession to decide whether we actually believe that the rights of copyright holders are worth defending. And if so, whether libraries themselves have a role in doing so. I’m now old enough to remember when we in libraries used to say things like, ‘librarians are the greatest defenders of copyright,’ but that's starting to feel like a very long time ago. I feel like now our profession is much more likely to conflate defending the integrity of copyright with defending big commercial publishing. With the result that, for example, we tend to get mealy-mouthed and equivocal about developments like the rise of massive piracy schemes like SciHub and LiveGen and Silent Librarian.

Now, that being said, I don’t want to give the impression that we in libraries have stopped talking about copyright or have stopped educating patrons about what it is and how it works. We do encourage patrons still, both to understand copyright and to follow the law. We spend a lot of time trying to help patrons understand how fair use works and how they can apply it and things like that.

However, these days, my impression is that we’re doing this much more as a way of helping our patrons stay out of trouble rather than as a principal defense of the rights of copyright holders, which I realize may kind of seem like splitting hairs. I mean, I can imagine somebody saying, ’look, we’re educating our users about copyright rules, and we’re encouraging them to follow those rules. Isn’t that the important thing?’ And I think the easy answer to that question is, yes, of course it’s fundamentally important to promote copyright compliance, regardless of whether we think copyright is a good thing because it’s the law, whether we like it or not. But I think the more complicated and the more accurate answer is that our underlying attitudes do matter because we communicate those attitudes to our patrons and to our colleagues, whether we mean to or not.

If you bring up SciHub or LiveGen or Anna’s Archive with any librarian, within seconds you’re going to know what that librarian really thinks about copyright policy, no matter what words they use in the conversation. And I’m very concerned that we have begun communicating to our patrons on the whole as a profession that copyright law represents something to get around rather than a set of rights that are held by actual people and that we should honor as a matter of principle. But to be clear, and I’ll just say this in conclusion, I also believe that honoring copyright strictly and exercising our fair use rights assertively are both important and that doing one not only doesn’t preclude doing the other, but also I would argue that both are equally important. We need to both help patrons understand what their rights are as users of copyrighted content and also encourage them not to infringe on the legal rights of others.

I think that this is a conversation that we need to have more openly and more rigorously in the profession sooner rather than later.

Thanks, Rick. I’m very challenged by your remarks. I remember writing once that fair use is not civil disobedience. And I still hold to that principle.

Rick Anderson

Yeah, and I couldn't I couldn’t agree more.

Kyle Courtney

So, in 2026 and beyond, what I’m watching for is whether the law finally acknowledges what libraries already know: that a knowledge system governed entirely by licenses cannot preserve access memory or the public interest without meaningful intervention.  And meaningful intervention could be a lot. Grassroots movements are great, top down movements are great, I will accept everything. There seems to be a big question about how licenses combined with technology block the library out of the digital space. The digital space is where our users are. So, will courts continue to treat licensing markets as dispositive? Or can we start to see some judicial recognition that copyright exceptions cannot simply be erased by contract? With the Internet Archive case, with the CDL case, with the AI training decisions, the pressure is building. We’re hearing it right now on this panel. Courts may be forced to confront a harder question: at what point does enforcing licenses undermine Congress’s design for copyright itself? Because copyright does include exclusive rights. But if you’ve ever read 101 through 106 rights, 107 through 122 are all the exceptions in which you do not have to pay, offer a license fee, or get permission—those are also part of the copyright law as well.

So, I think the Connecticut win will no longer be an outlier. I hope in 2026 what I’m asking for is whether other states to follow. We’re working with other states right now actively for this legislative session. Whether publishers challenge these laws, forcing courts to address preemption contract, commerce clause questions, all of those things that kind of dance around the copyright space, and whether legislators begin to explicitly frame library access as a public interest counterweight to the private licensing power that exists. Successful or unsuccessful challenges can be productive. As digital preservation starts to become a flashpoint, I think that’s where we’re going to see the rubber meet the road. As older licensed collections begin to disappear or platforms shut down or get gobbled up, we may see real irrevocable losses, not hypotheticals. So we need to act, because increasing pressure on fair use as a stable doctrine or these other things is very important, but there is a broader policy reckoning coming, which is ultimately what is ownership in the digital age. Too big for this panel, I would say, but a very big question that I’m hoping to get some more answers for in 2026.

Sara, take a look forward for us.

Sara Benson

I echo what Kyle has said about worrying about licensing overriding copyright exceptions. It’s been an issue that has been bothering me for a long time, and I agree it’s coming to a head, and we do need to preserve everything and can’t allow collections to disappear based on licensing.

Another thing that has been bothering me for a long time is DMCA exemptions for teaching purposes for online courses, especially with full videos. This problem persists, and we are again forced to reckon with the three year process for exemptions, trying to help our film professors and their students access full-length films for the purpose of teaching. This is maybe a “Sara problem,” but it’s one that’s been bothering me, and it’s a question that I get every single semester. When I ask my colleagues, almost every single one of them says yes. So, it’s a problem for academic research libraries.  

In addition to the DMCA problems, in addition to all of these other licensing [problems], I think as librarians we need to advocate, and we need to become familiar enough with the copyright laws and exceptions so that we can join organizations such as Library Futures run by Kyle and become advocates along with them. Because it’s not just the organizations that need to speak up, it’s the people who are at the front lines day to day doing the work. And so I hope everybody who is listening to this and or reading this brings it to their own library and their own staff and kind of gears up on these issues so that they can add their voice. Because we can go and we can advocate to the Copyright Office about the DMCA, but if it’s one person doing it, that’s not as impactful as if it’s 20,000 librarians writing it, right?

The ebook crisis is a crisis for academic and public libraries alike. And if you understand the issues a little bit better, that's when you can really advocate. So that’s my number one message: advocate for libraries. And, in order to do that, educate yourselves about the issues and then reach out to these organizations such as Library Futures so that you can get involved in it, and the American Library Association, of course, always advocating for libraries. That's my ending thought.  

Thank you, Sara. I would note that I raised these licensing issues in 2007 when I was working with the 108 Study Group, and it was summarily censored by the publishers. Jonathan, you get the last word.

Jonathan Band

Thanks. So, you’re hearing it first here that by the end of June, we’re going to have a clearer understanding of the bounds of secondary liability for copyright. And I say that with absolute certainty. Why? Because there’s a case pending before the Supreme Court, Cox v. Sony, and that’s the issue. The oral argument happened in early December, and the Supreme Court’s term ends at the end of June, and they will decide the case. So, unlike the rest of the speakers who were just speculating, I am offering you certainty that there will be a decision by the end of June. I don’t know what the decision will be. I can offer my thoughts, but that’s different.

Now, what is the issue? Oddly enough, the Copyright Act talks about infringement, but it never defines secondary infringement, meaning it talks about basically when, if you engage in the copying, are you an infringer? It doesn’t say anything about the person who assists the infringer and when they become liable. That’s all considered secondary liability, and there’s doctrines of contributory infringement, vicarious liability. None of that is specified in the Copyright Act, unlike the Patent Act, which has very clear rules for secondary liability. And the fact that the Copyright Act hasn’t had clear rules has led to a lot of confusion. There have already been several cases in the Supreme Court...we had the Sony Betamax decision in 1984. We had the Grokster case in 2005, and now there’s going to be the Cox-Sony case. And hopefully the Supreme Court will be much clearer than it was, meaning they were not at all clear in the Betamax case. They were clearer in the Grokster case, but evidently not clear enough. And so hopefully in this case, they will be absolutely clear about what the rules are with respect to contributory infringement.

This ties into everything libraries do, and it also ties into the issue of artificial intelligence, because when we were talking about artificial intelligence before, we were talking about the creation of the large language models, the scraping to create the AI tool. But a lot of the litigation and the harder issues, frankly, are going to be with outputs. When, if you use the AI to create something, at what point does the creation become infringing? To what extent is it substantially similar to earlier, what inputs were? And where this issue comes into play is what is the liability of the AI firm in outputs? If you have an output that is infringing, it’s pretty clear that if I as the user am inputting all of these prompts that create an infringing output, that lead to an infringing output, that I as the user, I would be the direct infringer. But then the question is, what is the liability of the device, of the AI maker, in that infringement? Are they direct infringers? Are they secondary infringers? Or are they not infringers at all?  

That question will be influenced in large measure by what the Supreme Court decides in Cox v. Sony because that’s where the court is going to hopefully come up with some clear rules about secondary liability. Also, Sara previously mentioned the Copyright Office reports. So the Copyright Office is working on a fourth report about this issue, about the liability for outputs. And I imagine they're going to hold off finishing that report until they see what the Supreme Court says in Cox v. Sony.

Jim Neal

Thanks, Jonathan. We will hold you to this.

That brings us to the end of our program. We have covered a wide variety of challenging copyright topics, and I feel like we could have gone for a week probing and debating the various issues that have been raised. I want to thank Library Journal for arranging this really important roundtable, and I want to thank the amazing members of this panel for sharing their expertise and their insights. That concludes our program.