Introductory Remarks by Lorena Guzmán-Diáz

Good afternoon, everyone, and welcome to the 2025 ASIL annual meeting, here at the Washington Hilton. It is my pleasure to introduce the highly anticipated panel discussion, “Digitized Brushstrokes: Are IP Laws Protecting Art in the AI Era?”

My name is Lorena Guzmán-Diáz, and it is my privilege and delight to serve as your moderator for today’s very interesting panel discussion. I will guide you through what promises to be a thought-provoking exploration of the intersection of art, technology, and law. Today you will hear from four distinguished speakers, each an expert in their field, as we dive into the complexities presented by artificial intelligence (AI)-generated art. Our session will unfold over forty-five to fifty minutes of discussion, followed by a ten-minute Q&A session where you will have the chance to engage with our panelists directly.

Before we delve into today’s program, allow me to introduce our four distinguished panelists whose expertise spans the spectrum of art law, copyright, technology, and the international legal scene.

Our panel will open with Pippa Loengard, Executive Director of the Kernochan Center for Law, Media, and the Arts at Columbia Law School. Pippa is a seasoned art law professor, and she will provide us with an overview of our current copyright law, as it applies to AI-generated art, along with the challenges and legal disputes that have arisen in this new frontier.

Following Pippa, we will hear from Jim Rosenfeld, who brings his legal expertise on media, IP, and internet law to the table, including specialized knowledge on the DMCA and the CDA’s legal immunities concerning user-generated content. Jim’s insights into AI’s technicalities and content creation and its implication for our legal frameworks is truly going to be invaluable.

Next, we will hear from Maria Vazquez, Dean of the Law School at Universidad de San Andrés and Director of WIPO’s Latin American Program in Intellectual Property and Innovation. She will share her international perspective. Her experience in music law and her work on copyright and emerging technologies will shed light on the various global approaches to AI art and its impact on local laws and markets, particularly in Latin America.

And then finally, Joshua Kaufman, Partner at Venable and Chair of their Art, Copyright, and Licensing Group will discuss the dynamics of the art market and valuation in the context of AI artwork. With Joshua’s deep experience in art and copyright matters, we will gain insights into the role of authenticity and provenance in the digital age.

Each session promises to offer groundbreaking insights and to foster dialogue about the future of artistry, ownership, and the law in the AI era. In closing, we will summarize the key takeaways from each segment and transition to a Q&A session where our audience can participate in this vital dialogue.

Let us set the stage for an enlightening journey through the burgeoning field of AI art, its legal intricacies, and the evolving definitions of originality and creation. Please join me in welcoming our first speaker, Pippa Loengard, as we commence this exploration as to whether IP laws are indeed keeping pace with the digitized brushstrokes of artificial intelligence. Thank you all for joining us, and I am happy to begin.

Remarks by Philippa Loengard

Thank you so much, Lorena. Thank you for having us here, and to all the organizers, thank you so much for this kind invitation. Not setting it up for failure or anything, groundbreaking.

How many people in this room have a background in intellectual property (IP)? Okay. Great. I am going to start back at the beginning, and so folks who are new to the area will be caught up by the time I am done.

Lorena and I were talking, and one of the first questions that she posed is what are the disputes and challenges to traditional copyright? The first thing is that this is not new. We have been down this road many times with new advances in technology and various iterations of those new advances.

The first case that is important is from 1882, which was when the Supreme Court decided that photographs, at that point in new technology, were copyrightable. The determination was based on the fact that the photographer takes great care in the selection and arrangement of the subject. Those two words, I think everybody on this panel is probably going to talk about at some point. We have proceeded 150 years from then, but we still use those words, and they are important in the AI story.

The problem is that selection and arrangement can cut both ways. The same way that selection and arrangement allowed for the protection of photographs, we are going to discuss later how that can curtail the copyrightability of artwork. I am also going to make one point, and that is that copyrightability is important because without it, a copyright holder in the United States cannot sue. When I talk about copyrightability, I am talking really about the right of the owner to participate in our legal system. Sometimes there are suits over various things, but a lot of the controversies that we are going to talk about today are just about copyrightability itself.

Another case that questioned photographs, granted another 125 years later, was our dear friend, Naruto. Some of you may know him. He is a portrait photographer specializing in selfies, and it was determined that while a separate photographer, David Slater, had set up the camera, Naruto pushing the button meant that no one could have copyright in that work because there was no human authorship. This was ten to fifteen years ago. But the point being is that this idea of human authorship has reared its ugly head for quite some time. We have problems with human authorship throughout the period, throughout art.

We have studios, for instance, which is not a human authorship issue. Someone is authoring it. But the question of who has always been sort of up for debate. Do we need a specific author? There are work for hire issues there, which I am not going to go into now. But this whole idea of who is an author is not new. This is a picture of a garden that was deemed by the Seventh Circuit not to be copyrightable because nature is the author of a garden.

We have Felix Gonzalez-Torres’s “The Kiss.” This artist did not seek copyright. But this was a situation where there would be a pile of specific candies in a corner, and visitors were encouraged to take a candy, which meant that everyone altered the sculpture in some way. So is that who was the author there? Now there are multiple humans.

We have one of the earliest examples, which is from the 1950s, which is the Méta-Matic by Jean Tinguely, which was a situation where you could put a marker into a clip. There was a pedal on the floor that you hit, and the machine would do the rest. The machine in its operation was naturally jiggly. It was meant to be. The paper itself was not static, and so every piece created was new. Who was the author there? Was that the author, the marker, not a human, the person who hit the floor button, or was it Jean Tinguely who had set up the machine to begin with? This is not new at all.

None of these cases have looked for copyright. In fact, in the Naruto case, it was the People for the Ethical Treatment of Animals that sought copyright, not Slater. Going back, we generally do not encounter this. What is interesting with AI is that we are encountering it because large corporations are at stake. They are part of the process. Any time you have that much money and power involved, we are going to have more coverage and controversy.

For the art, on the artist front, the issue right now is on the training data that is used, the scraping of the training data. That is one issue. Then the second is, again, the copyrightability of the works that they produce with these platforms through these algorithms. We can get to that whenever you would like.

Lorena Guzmán-Diáz

I know you are saying that these cases are not addressing copyright. Do you think that there is a need for any specialized legislation for art generated through AI, and what might that look like?

Philippa Loengard

We have this issue between inputs and outputs, and we have two artists on each end. We have the artist whose work was on the internet and scraped and put into the algorithm, and they are looking for transparency legislation because the way that they can bring a suit in federal court that their copyrights have been violated is to know that their works were part of the algorithm, were used, were scraped from the internet, made into the algorithm, and thus a new work could have a basis in their work. That is part one for artists.

Part two is artists who are creating in this sphere, and we have some great artists who are using this technology and generating interesting and ever-changing works.

This is a piece that is called the “Entrance to Paradise.” This was an artist named Stephen Thaler. He is suing for copyrightability. It is a piece that was generated purely from AI. He has lost. He is actually the only current suit. What these folks are looking for is not so much legislation but recognition from the Copyright Office that their works are copyrightable. Legislation on the output side has not become an issue yet.

Lorena Guzmán-Diáz

Can you provide some other examples of disputes that have challenged our traditional understanding of copyright?

Philippa Loengard

We have always thought of copyright as something that protects the products of humans, and we say that not because we made it up, although you could say that everything is a legal construct, but that it is defined by the Copyright Office in what they call the compendium, which is an interpretation of the law of copyright, where they say that the Copyright Office will only register works of authorship provided it is work made by a human being. They cite an 1879 case about original intellectual conceptions of the author.

In the Thaler case, the District Court again noted that it has to be a human because all of copyright is based on humanity in the sense that the copyright term is based on the life of the author plus seventy years. That implies that there is a person who can grant their rights in a will to their heirs. Only humans have heirs, that in order to register, you need a signature of the author and only humans have signatures. These are implied ideas of authorship.

It is more complicated now. Because it is not only who is the author but who is an author and the whole concept of authorship in general and what we expect as people, as consumers and as artists, what we expect out of the term “author.”

One of the issues that we have in authorship is this idea of style. That is a tangent in a way, but I think it is coming up, and it is going to come up even more because a lot of what you get through algorithms is something in the style of an original author. If you ask any of these programs, Midjourney, etc., to create irises in purples and blues and yellows in a vase on a table, you might very well get something in the style of Van Gogh. Van Gogh’s works are not under copyright at the moment, but there are plenty of artists whose works are, where you could get a similar style.

One of the most current artists who is leveraging this idea of style being stolen, if you will, by AI and the abilities that AI has is Hayao Miyazaki, who some of you may know. He is a very famous anime creator. Ironically, he created a character named Naruto too, but this is some of his work. A movie was recently made, a trailer for “Lord of the Rings,” obviously not related to anime at all, but in the style of Studio Ghibli, which is Miyazaki’s corporation. And no bones about it, totally took it off.

It is clearly shockingly similar to Miyazaki’s work. While you cannot copyright style, you can trademark some of the style items in his work, and this seems to me like a pretty clear violation that would be more easily than others proven in court to have taken specific attributes that were copyrightable, and it could supplant the market of Miyazaki’s work because it is so similar. If you put it in a situation, not a film trailer for “Lord of the Rings,” but say an anime situation, you would have a real problem.

Lorena Guzmán-Diáz

In that same vein, what role do you believe that the artist programmer should play in the legal recognition of AI-generated works?

Philippa Loengard

That is a tough question. I am not quite sure. The artist is always going to be the person who is spurring on the creation. They do it in various ways with various levels of participation. This is a work by Kris Kashtanova that was submitted to the Copyright Office and mostly rejected except for her initial design, which is the drawing on the left. The drawing on the right took 265 prompts to get what she wanted out of the AI. What role does she play? She played a really important role there. That said, she was denied copyright for most of the work, and this is a comic that she did.

This is a pretty self-descriptive title of “Slice of American Cheese.” That is sarcasm. I have no idea why it is called that. This actually did get some level of copyright because of the selection and arrangement of the work, and what this artist did that Kashtanova did not do was to submit a video of him doing the prompts over and over and over again.

Is that what we are going to coming to? The more videos you can upload to prove how much you were involved? I do not know.

Lorena Guzmán-Diáz

The more prompts.

Philippa Loengard

The more prompts I see. The more that you guide the AI into giving what you want, I see that. How we start proving that, I am not sure.

Lorena Guzmán-Diáz

Thank you so much for your comments today.

Now I want to switch over to Jim, who is going to talk to us about the technological perspectives and solutions. Jim, from a technological standpoint, what are the unique challenges that AI art poses to IP law in your viewpoint?

Remarks by James Rosenfeld

The challenges mostly arise from all of these issues of ownership and authorship that Pippa was discussing. Because AI-created works are not copyrightable, there is going to be a huge corpus of works out there that you would see and think were creative, original works that are not copyrighted and not copyrightable. Many works may not have authors or it might be difficult to determine whether a work has an author or determine what elements of the work are protected by copyright, and that poses a lot of questions, both for people who are trying to protect their rights and people who are trying to defend their rights.

It will be more difficult for creators of AI-generated works to protect their rights because they may not own the entire work. If you did the prompts, you might own the selection and arrangement, like in the case of the American cheese case that we were just talking about, but not the components that are selected and arranged.

You might be something like a joint author, although you cannot really be a joint author with IP, nor can you create a derivative work from an AI-created work. That is going to become more difficult and complicated to prevent copying of works and to go after infringers.

Lorena Guzmán-Diáz

As we consider AI either as a tool or as a collaborator, what are the potential liabilities for content creators relying on AI?

James Rosenfeld

Publishers and artists are suing AI companies. the New York Times, other publications, groups of artists, groups of writers. There are about thirty cases pending in the courts in the United States. They are also able to and will go after the end users of this AI, not just the AI companies themselves, but anybody who is using the output for anything. If you are a creator who is using AI as a tool, you might be the subject of copyright claims from somebody, from the New York Times or whoever claims my work was input and your work came out, and to some extent you are copying me, or this element of your work is copying me. There is greater risk there that they will eventually be the targets of these claims too. The pending cases may determine whether that is copyright infringement or fair use.

Perhaps, in some ways, it is more safe for creators to use AI-generated work that does not have an owner than to use another work that does have an owner. As Pippa was saying, it gives them a lot of ability to do a lot of interesting things, and so I think there will be a lot of benefits of AI to creators as well.

Lorena Guzmán-Diáz

I know a lot of your expertise is on the DMCA and the CDA, and I was hoping you could discuss the applicability of both of those in cases involving AI-generated art, especially regarding user-generated content, maybe what you have seen in your day-to-day.

James Rosenfeld

This is all theoretical because these cases do not exist yet. We are in early days. The application of Section 230 and the DMCA Safe Harbor to AI just has not played out yet in the courts in the United States.

But let me take one step back because I know this is a group of attendees from all over, and I want to make sure everybody understands what we are talking about. Section 230 and the DMCA Safe Harbor are the two main U.S. statutes to protect websites, platforms from liability for content of others that appears on their websites. And 230 protects against liability when somebody is seeking to hold them liable, say, for a defamation claim or write a publicity claim or a false advertising claim based on content that somebody else has put on their website or service. The DMCA Safe Harbor essentially serves that function in the copyright world. If you are YouTube, it protects you so long as you do not have knowledge of an infringing video that appears on your platform.

For the most part, these two legal protections are going to continue to operate the same as they have and would apply the same to AI-created works as other works in that 230 will still protect you if you are hosting defamatory content, whether it is created by a computer or a person. The DMCA will still protect you from infringement liability, again, assuming you do not have knowledge of the infringement, whether it is an AI-created work or a human-created work.

But I think there are some more difficult issues on the input side as opposed to the output side, at least with the DMCA Safe Harbor, because as I mentioned before, creators may want to assert claims against users of AI. They may want to assert claims against platforms on which AI-generated works appear. I think it is going to be very difficult for those users to detect in any one work when their work has been used. How can you tell which work has been used in a certain piece of output? It is going to be very difficult to find that. It is going to be difficult to find whether the work is in the database that was used to train that large language model. There are going to be difficult issues that are playing out in the courts now about whether that is infringing or fair use.

Even if a copyright owner is able to establish liability, what is the relief there? Is the relief to remove that work from the corpus that has already been used to train the AI? I think technically that is very difficult to do, and in some sense, the horse is out of the barn because it has already been trained. There may be complications that arise in that sense.

Lorena Guzmán-Diáz

In your experience, maybe in your practice, have there been any precedents where DMCA takedown notices have been effectively used or challenged with AI-generated content?

James Rosenfeld

Not yet. I am sure it will lead to a multiplying of takedown notices, just the way it has led to this huge hassle of litigation in the United States. We have been talking about a lot of difficult issues about what is and what is not infringing. It might be that platforms get a little more risk-averse with work that they know is AI-created because they think, well, I am better protected here because this does not have a copyright owner or very little of the content is protected or it is fair use. But I think for the most part, platforms will continue to be cautious and take things down when they get takedown notice because that is the safest approach.

Lorena Guzmán-Diáz

I want to turn to Maria Vazquez, who is going to talk to us about the international and local approaches to AI art.

Remarks by Maria L. Vazquez

Thank you so much, Lorena, for moderating this panel, for inviting me, and many thanks to the American Society of International Law. It is my first time in this association, so it is really lovely to be here. And it is a long way from Buenos Aires. It is always a pleasure to get to D.C.

Lorena Guzmán-Diáz

How do approaches to IP protection for AI art differ internationally?

Maria L. Vazquez

Pippa and Jim have already spoken about the two big issues when we talk about AI and copyright, and that is the input and the output, and hopefully, we will be able to talk about the input in a bit because I believe that this is where the approaches do differ more with respect to the output. That is, whether or not a piece of artwork created by AI is actually protectable by copyright.

I believe that, yes, the approaches may differ a bit, but not dramatically because basically copyright all over the world is based on a premise of human authorship. Even though some legislations actually say this, others do not, but it is always somewhat implied.

Specifically, I would say that the U.S. Copyright Office, as Pippa said, has already pronounced itself in this regard, saying that there has to be at least an element of human authorship. In Europe and Latin America, which tends to follow Europe in everything related to civil law and to IP law, we are all signatories of the Berne Convention. And the Berne Convention is somewhat based on a premise of human authorship.

I love turning back to an article by Jane Ginsburg, who in fact is from Colombia, where Pippa is, and I really love that she talks about people, not machines. Coming back, I think a bit of a play on words on an older Sam Ricketson article—and I do often teach this when I am in class—in the early 1990s, he would talk about people or machines but always coming to the conclusion that human authorship is the premise in the Berne Convention. Professor Ginsburg does the same when she talks about people, not machines. I do think we all tend to lean toward this same decision.

However, the UK has a bit of a different approach because they do speak in their law of 1988. They do speak of computer-generated works. But even in that case, they offer a certain protection, but it is a more moderate protection, different in the number of years of protection and its duration.

We have seen a different approach from China. China is not entirely clear in the five or six cases that I know about. I am not an expert in Chinese IP, but apparently, there is not as much respect for legal precedent. They seem to come out with somewhat diverse outcomes.

But very clearly in the second case I believe it was 10 percent against Shanghai Ying Shan, and that case says that AI art will be protected as long as it has human input. But the human input, as long as it is prompt engineering, that would be enough. It is quite a bit more flexible than many of our other jurisdictions.

I do not know if this is really the case, but I am going to take a leap and say that we have seen it in patent law, how, in fact, the WIPO says that China is the leader in patent applications, not only in China, but also in PCT files for international patents. This has happened in the last two decades. I see that in copyright, there is also more flexibility than in other countries.

When filing a patent, they have become more flexible in registering patents in the last two decades, and I would maybe venture to say that in the last years, for example, in 2023 in China, there are about 46 percent more cases of copyright works registered than previously. The same in 2024. There are more works registered. Perhaps this has to do with more flexibility toward AI artworks, toward that protection.

Lorena Guzmán-Diáz

Would you say that this is one of the lessons that can be learned from the way that other jurisdictions have approached IP policies, and that they are adapting to these technological advancements in arts or do you have any other lessons learned?

Maria L. Vazquez

It is hard to think about lessons learned when all of our countries are trying to see how to adapt to this new technology.

When it comes to input, what I think is that in places other than the United States, in the European Union, in Latin America, we do not have fair use as part of our copyright law. This is quite well known. We actually have exceptions, a very closed catalog of exceptions, very specific exceptions.

Why is this important for those of you who maybe are not in copyright or do not work in copyright? Exceptions in international legislation have to do with instances where one can use someone else’s artwork without asking for permission. But, as I say, these are very precise, such as in an academic written work, such as a thesis in college. Yes, you are allowed to quote another author without actually calling him up and asking for his or her permission. This is the right of quotation in very specific works, and as such, there is also a right to use a work without asking permission in academic use and in educational uses. These are the kinds of exceptions that we have. Some people have library uses, some countries.

Fair use is very different from these very specific and closed catalog of exceptions. Fair use has to do with these factors or criteria that judges apply in different ways, and so as a copyright professor, I teach copyright in, as Lorena mentioned, an IP program for all of Latin Americans who actually come to Buenos Aires to take these courses. I do always teach copyright and derecho de autor, droit d’auteur. I teach the differences. I am a big fan of the copyright system and of the Fair Use Institute as such.

Fair use cases in our program are always a big hit because they are all different. They are so colorful and so unpredictable. I just love teaching those cases where you put it all out there and everyone is waiting for the outcome, and then they say, what do you mean, Patrick Carriou did not win against Richard Prince? When he used it and all he did was paint. No, that was considered actually a transformative use because it is a new message. Or what do you mean you can copyright a vibe in a song? Is that not like an idea? Does that not mean that ideas are protected? It was considered fair use.

Here we are now, all waiting, the rest of the world, waiting to see what is going to happen in these cases that Jim so eloquently spoke about, the thirty or so cases that are all in the United States. There are some in Europe. In Latin America, we do not have any of these. Here we are, all waiting to see whether this will be considered fair use or not, because this will have a big impact in the rest of the world, what happens in these U.S. cases. We are all waiting to see what will happen, because later, when our artists want to maybe negotiate a license with these big AI companies, we have to see how much negotiating power they have, and perhaps their negotiating power will depend on what comes out of these cases.

Lorena Guzmán-Diáz

Speaking of the impact that these cases in the United States are going to have, do you see any potential for the harmonization of laws concerning AI-generated works?

Maria L. Vazquez

It would be wonderful if we had a treaty. In fact, whenever I go to Geneva, I always suggest it: Can we start work on an AI and IP treaty? We could all get together, this whole panel, and people from all over the world and see if we can get together something. One of my graduates just very recently published an article in the Yale Law Journal. She is really excellent, Micaela Mantegna. I, in fact, was able to mentor her thesis, and she published it as a book. It is one of the first books in the Spanish language that talks exactly about these issues. And in the Yale Law Journal, she says perhaps Gen-AI should not be viewed or perhaps copyright law should not be viewed as such an important element in Gen-AI.

Why does she say this? Because she says it has to do with so many laws, with data protection, with consumer law, with human rights. It has to do with many other things. The reason she says this is because we are up against very big and powerful companies, as Pippa mentioned.

Perhaps in this AI treaty in WIPO, we could bring together different aspects of the law, different areas of the law. That would be wonderful. I do not see it as very plausible, because in fact, what are the big challenges? Approaches do differ from one country to another, on the one hand, and the fact that there are so many interests at stake, because we do have this idea of protection of our works, of the works that are being imported or being scraped off the internet, as input for AI systems.

Even in countries such as mine, in developing countries, in lower- and middle-income countries, we also want our scientists, our people in technology to be able to innovate. If in our national congresses we are not able to harmonize opinions and make an AI law, imagine between all the countries, a harmonization. I would definitely suggest it. I, in fact, have also suggested it in PROSUR, which is a Latin American association of about eight countries, I believe. Could we do something like that? I think it would be good.

But when you are balancing interests of stakeholders, such as artists, on the one hand, and content producers, on the one hand, not just artists, but The New York Times, authors, et cetera, and on the other hand, innovators. As I say, in our country, it is very hard to lean totally toward copyright, because we do not want to limit innovators in that they cannot compete later with the Global North or with China. We want our innovators to be able to create AI systems that work. So it is really not easy, is it?

Lorena Guzmán-Diáz

Thank you so much for your comments today.

I want to switch it or turn it over to Joshua Kaufman, who is going to talk to us about art market dynamics and valuation. I do not know if you have a few comments.

Remarks by Joshua Kaufman

First of all, Lorena left off the most important part of my CV. She was my student in art law, so that is my claim to fame.

A couple weeks ago, Christie’s had an auction of AI art. 6,500 artists signed a petition opposing it. They went ahead anyway. They sold thirty-four lots, mostly to Gen X and Millennials. There were a few high prices, and the rest were not. But it reminded me, almost four years to the day of that auction, there was another auction at Christie’s of digital art, NFTs, 69 million. I would say we are probably talking about 69 cents worth of value today on those. It is a piece of paper. It is a piece of canvas. It is a rock that somebody has scribbled on. There is no inherent value in the materials. Art gains its value by perception, by what a community thinks of it on a particular day. It changes. People whose was valued in astronomical sums is worth very little today. People who are disregarded. Norman Rockwell in the past, he is just a dumb illustrator. Today he is a great fine artist, and his works are going for a million. It all changes. It is all based on perception.

I am going to give you a small story. Years ago, I was representing Joe Hirshhorn, and when he first moved from the museum, when he first moved to D.C., he wanted to get a sense of the local community. He went to the Washington Project for the Arts, a tiny little 501(c)(3). And there are the ugliest things there I have ever seen in my life, these candelabras, they are dayglow pink. They are kind of like a woman’s shape, so where the breasts would be, they put pasties and glitter. They were so ugly, you could not imagine it. He buys them for 50 bucks.

Fast forward three years later. We are at the annual event at the Corcoran. It was a wonderful museum that is no longer around, and everybody is dressed to the hilt. They bring an auctioneer down from Christie’s or Sotheby’s. There is an auction. Joe and Olga, his wife, are the honorary auctioneers, and an identical pair of these ugly candlesticks come up. Not Joe’s, another ugly pair, and the auctioneers starts, “500. Do I hear 500? Do I hear 450? Do I hear 450?” He is down to $75. People are laughing. Dear Olga, who never broke five feet, takes the mic, brings it down to her four-foot-eleven height and says, “I just want everyone to know there’s another pair of these in Joe Hirshhorn’s personal collection.” Ninety seconds later, they hammered for $1,500, What no one would pay for, not 50 bucks, as soon as they found out it was in Joe’s collection, they were worth $1,500.

So when you look at art and value, there are so many indescribable elements of it. One of the things that makes art valuable, though, is scarcity. Theoretically, a one-of-a-kind is going to go more than for a limited edition print, which is going to go more than for a larger limited edition, which is going to go for the smaller. And they all go for much more than a poster edition.

If you have no copyright protection on an AI-generated artwork, anybody can knock it off. There is no scarcity. On a limited edition print you put in the size of the edition, and it is locked in. Anybody can take a picture of one of Pippa’s slides and make a million copies. Where is the value in something when there is no scarcity? Is it the piece of paper, the piece of canvas? No, there is no value there. You have a real inherent problem of having any value attaching to something that can be made by anybody in any format. You want to make a jigsaw puzzle? You want to make a roll of toilet paper with it on it? That is okay. You can do anything you want with the image. Where are you going to build value in that?

Art licensing is an $8 billion-a-year business in the United States; it is huge. It is everything from greeting cards to postcards to this carpet to your blouse. Everything is art license. But in every art license contract, you have to warrant that you are granting an exclusive, at least to that image in that category. I grant you the exclusive rights for carpets. I grant you the exclusive rights for jigsaw puzzles. You cannot grant exclusive rights to anything that anybody can copy. What manufacturer is going to want to license your artwork for greeting cards or postcards or whatever if somebody else can be a competitor right next door and you cannot do anything about it? You cannot warrant what you are doing. You cannot limit anything else.

Then as to authorship, who owns it? Are you, as the person who is typing in the prompts, going to go to the person and try to sell it and say you are the owner? Back to the gorilla case, who is actually the author and the owner of it?

Lorena Guzmán-Diáz

How do issues of reproduction rights and artistic royalties come into play with art that is generated by AI?

Joshua Kaufman

Royalties are basically you license something and you get a percentage of the wholesale value of it or the retail value. If you cannot license things, there is not going to be much of a royalty. Then again, who is going to claim the rights? Can the programmer claim rights? Can ChatGPT claim part of the ownership? Is the person who is putting in the non-protectable prompts? Prompts are generally ideas. You have to remember instructions are not protected, recipes are not protected, how-to is not protected under copyright. So ideas are not protected.

What is a prompt? You can put thirty different unprotectable things together, and they are still unprotectable. I do not think it is going to be a big issue with licensing and royalty splits, because I do not see a market. I do not see anybody paying anybody anything for something that everybody can copy.

Lorena Guzmán-Diáz

As an art collector, what kind of advice would you give to artists or to art collectors when dealing with contracts or authentication of AI-generated art?

Joshua Kaufman

I have written a book on the art of investing in art, and I lecture on it—only buy art that you love at a price point that you are comfortable with hanging on your wall. Art, by and large, is a horrible investment. You read all these things in the newspaper, but if you look at 990,000 of the 10 million pieces, they depreciate. You go to a gallery; you buy a print for $1,500. You are going to find it on eBay for $250. You try to sell a $10,000 piece that you bought. You go back to the gallery. They are not the least bit interested in it. No museum will take it as a donation.

Buy art because you love the object and you want to live with it, and you can afford the price, period. That is my advice on art investment generally.

You are going to have a problem with authenticating. If these are digital files, digital prints, you do not know where they came from because anybody can duplicate perfectly a digital file. It is not like in the old days you had first and second generation, where you had the record, then you had the cassette, etc. There is no deterioration in quality if you get a good scan or you get access to the original file. There is not going to be any difference between your knockoff and your original. It is not like your Coach bag where they use lousier leather. It is going to be identical. When it comes to authenticity, yes, one can come from the artist.

A quick story. I represent David Hockney and he has all of these video projections that he does, thirty-six screens and things like that. When we sell those or donate them to museums, the museum or the collector gets a box. It has an SD card in it. It has a DVD of the whole program and a hard drive that is a backup of what is on the SD card. And then you have all of the monitors that people had to buy on their own in the computers. Where is the artwork? What has the value? We said it was the SD card because that is what David created. That is what downloaded from his computer. The rest is all ephemera. The SD cards, what were numbered and what were signed by David, that is the artwork. It is not what you see on the screens. That is not the artwork. The artwork, from the art world point of view, is the SD card.

Warhol’s Brillo Boxes and all those things you have seen, he made the first set, and then he authorized people to make the other sets. If Warhol did not touch it, they are not Warhol. All of those other Brillo boxes that had been Warhols for generations, that sold for millions, all of a sudden became worthless, because even though Warhol authorized it and they were exact replicas of the ones he touched, he did not touch them, they had no value.

You are creating a digital file that somebody else is reproducing, no value.

Lorena Guzmán-Diáz

Thank you so much. I want to thank all our speakers today for all their comments.