Episode 20 - Is AI the end of Intellectual Property?

AI is increasingly viewed as a potential co-creator of text, images, videos, music, and audio.  However, in March of 2023 the US Copyright Office expressed the view that ‘copyright can protect only material that is the product of human creativity.’

In this episode of Exploring AI Matters Claudia Ray, a partner in the New York office of Kirkland & Ellis, a major law firm with a distinguished IP practice, discusses the important questions of intellectual property raised by the newly emerging artificial intelligence technologies, particularly large language models like OpenAI’s ChatGPT and Google’s Bard.

Transcript

SPEAKER_02 0:01

Welcome to Exploring AI Matters. This podcast series, previously known as Mind the Gap Dialogues on Artificial Intelligence, will continue to appear in the ABA series to the extent that, in addition, all of the episodes, old and new, will now appear under our new podcast name, Exploring AI Matters. Thank you.

SPEAKER_06 0:50

Today, three and a half years later, we revisit intellectual property and AI. In the intervening time, AI has become a global topic of discussion. In March of 2023, the U.S. Copyright Office expressed the view that copyright can protect only material that is the product of human creativity. Moreover, one of the issues raised by the screenwriters and actors guilds in their recent strikes was the prospect of generative AI being used by studios to bypass their services. Welcome to Mind the Gap, Dialogues on Artificial Intelligence. I'm Ama Adams, a national security lawyer.

SPEAKER_04 1:26

And I am Charles Palmer, a computer scientist. We are your hosts for this episode of Mind the Gap Dialogues on Artificial Intelligence. In addition, we have two more hosts.

SPEAKER_02 1:38

Hello, I'm Roland Trope, a national security lawyer.

SPEAKER_05 1:42

And I'm Mark Donner, another computer scientist.

SPEAKER_06 1:45

Each episode will be led by two of us, with others adding impromptu questions and comments as the spirit moves them. Our guest today is Claudia Ray, a partner in the New York office of Kirkland and Ellis, a leading law firm with a distinguished IP practice. She represents clients across industries, including entertainment and media, financial services, technology, and consumer products. Claudia is an active member of several professional legal bodies, including the International Trademark Association's Bulletins Committee, and she previously chaired the New York City Bar Association's Copyright Committee. She frequently lectures and publishes on IP-related issues.

SPEAKER_04 2:25

Thank you for joining us today, Claudia. For the non-experts in our audience, which very much includes me, would you briefly explain copyrights and why they are part of our law?

SPEAKER_00 2:37

Sure. So copyright is the body of law that protects what we call original works of authorship. And it is a type of intellectual property protection. The idea of copyright protection is set out in the Constitution. It's Article 1, Section 8, Clause 8, that authorizes Congress to promote the progress of science and the useful arts by granting authors who are creators of original works the exclusive right to their works for a limited time period. So that's where we get the basis of copyright law. And we have several statutes that have been enacted to implement that constitutional authorization. The most recent one that we all operate under today is the Copyright Act of 1976, which actually took effect in 1978. And it provides that authors who create original works have various exclusive rights. Section 106 of the Copyright Act gives them the exclusive right to copy their works, to publicly perform them, to create what are called derivative works based on their works, to do various things with them. That protection is actually quite long term. Currently, for a work created by an individual, copyright protection lasts for the life of the author plus 70 years. And for a work that is created as what we would think of as something done by a company through its employees or as a work for hire by a contractor, the protection lasts for 95 years from publication or 120 years from creation. So this is very long-term protection as compared to things like patent protection, which is much shorter. And we've had a lot of debate over the years about whether it should last that long. But currently in the US, it's quite a significant period of protection. And the idea is that there's this economic trade-off that is baked into copyright law. That as the Constitution says, we want to promote the progress of science and the useful arts. And the idea is that the way to do that is to say to authors, well, if you create something, you will get this period where you're the only one who can exploit it. And you know then that you will have some economic benefit from it. Nobody else can sell it, nobody else can license it. And so you'll be able to generate revenue from it to support yourself and to support the creation of additional creative works. And so here's the trade-off. We're going to incentivize you with these exclusive rights, but we also recognize that it benefits society to have these things available. So eventually the protection will end. And so that's the trade-off that we have baked into the Constitution and baked into the statute. And the idea is you're going to have the creation of works continually happening because there is this promise of an economic return to the creators.

SPEAKER_04 5:33

Well, that's that's quite a quite a broad area there. Well, anyway. Can you discuss how these how today's digital technologies have messed this up? Are they disrupting copyrights, either breaking assumptions or challenging established business models?

SPEAKER_00 5:52

Sure. And and we've seen sort of different waves of digital technologies. When you think about copyright, it's really quite a broad subject matter, right? It's it's not patentable inventions, it's not trademarks, but it is many, many other things. So copyright could cover a musical work, for example, both the underlying composition and a sound recording. It could cover the text of a book, it could cover a motion picture, it could cover both the actual audio-visual work that constitutes the motion picture and the underlying script for it. It could cover a play, it could cover a sculpture, architectural drawings, it can cover dance notation that is used to fix a dance in a tangible form. So there are all kinds of things that can be covered. And what we've seen is that there have been different waves of technology that have been disruptive. You can go back in time to player pianos, and people talked about that at the time as disruptive. But currently, we think about disruption in connection with digital technologies, in connection with things really involving music as one good example, right? When we first had the advent of digital downloading and digital streaming, that was really disruptive for the market for recorded music, which had been sold in various formats. Originally we had LPs, you know, physical albums, then you had CDs, you had audio tapes, and then you moved to the digital format and you had things really changing over time. So you had the unbundling of albums, for example, where you used to have to buy the whole album, and now you could just buy one song and it could be downloaded to a device where you didn't have to have any physical medium where that work would reside. We also have streaming technology where you can listen to music not over a radio, but through a device or through your computer. And that really changed things about how people listen to music, about how music is used and enjoyed, but also how the people who create the music, and that could be the songwriter, it could be the record label, it could be the performing artist, it could be a venue, all kinds of folks who are involved in the creation and delivery of that music, saw that the model for them to be compensated to get that economic return from those copyrightable works was really changing. And initially there was a lot of concern about how are people going to enjoy these works. And you saw music being shared online in digital formats where people were sharing it without paying the original authors. And there was a lot of concern about that. And we've had the development now of streaming platforms as just one example of ways in which people enjoy digital music, and the original authors can actually get paid for that. And there are debates about whether they get paid enough and whether the model actually works. But we do have a model today that does involve payment for music as opposed to people just sharing digital files of music online with no expectation there will be payment. So it's taken time for that to work through. And some people would say it's still working through and it's never yet been perfected and that it still has a way to go. But we have worked through and we've gone a ways from the early days of that kind of digital disruption in the music industry, is just as one example.

SPEAKER_04 9:16

Yes, I remember when all these things just got started. But speaking of getting started, AI is here. Technologies, including large language models, chat GPT, and all the derivatives. These are their latest disruptors, I suppose. What what sort of issues are these raising? Is it new?

SPEAKER_00 9:35

Well, you know, it's really interesting. And we're seeing, I think a lot of folks sort of grappling with what does this mean. We see clients who are trying to think about ways that they might benefit from AI, that they might be able to integrate it into their business and automate some of the things that they do using AI. And folks are really thinking about it, grappling with it, trying to decide is this a help, is it a hindrance? You're seeing a lot of issues for users who are adjusting to these new technologies and sometimes with an imperfect understanding of what they do. You are seeing even in the legal profession, people are grappling with AI and can it make us more efficient and better at what we do? And then at the same time, you see folks really struggling to understand it. And in some instances, using things like ChatGPT to do what they think is a search for case law and realizing that in fact what they're getting are sort of fake results in a sense, where the results that they get back are not cases that have actually been issued from a court, but rather things that the AI is essentially making up, so that they really are hallucinations of the AI and running into problems with that when those cases that don't really exist are getting cited in papers that are being filed with a court. You also see, as you know, in the the recent Screen Actors Guild strike, a lot of concern around AI being used to replace human actors and performers, where people in crowd scenes might be generated by AI rather than have extras getting that work. So you're seeing all kinds of sectors of our economy and economies around the world where there's an impact. And I think we're still in such early days that nobody really knows how that's all going to shake out, but there's a lot of interest, but at the same time a lot of concern as well. And it's it's certainly an interesting time to be involved with these issues and involved with copyright and thinking about this.

SPEAKER_06 11:34

And Claudia, I mean, you certainly hit on a point because you're kind of walking through the different issues that companies and users are thinking about, right? Like AI is sort of all the rage, everyone's talking about it across industries. And I, you know, from my vantage point, sort of the issues and the concerns seem to be a little bit all over the place. I suppose that's depending on how AI might be impacting an individual user or business or how leaders or management are thinking about that. But you know, based on your practice and sort of the breadth and scope of your practice and the types of clients that you're talking to about AI-related issues in their industry, you know, I think that gives you a really unique vantage point or sort of visibility into what is keeping business and tech leaders up at night. Like what are the top issues that they are thinking through as they sort of grapple with AI and what that could bring?

SPEAKER_00 12:29

So there are a couple of things that we're seeing clients talk about. And I don't think these are unique to our clients. I think folks generally who are interested in AI or dealing with AI or thinking about this. But one issue is what is the ownership status of a work that's created using AI or is created by an AI? How do you deal with that? If you're in a creative field and you're using AI to make the output, is the business then going to own something that it can protect? Is that output going to be protectable by copyright or some other form of intellectual property protection? Is it going to be an asset or is it going to be something that perhaps isn't protectable? And do you then devote resources to creating this thing if there's no protection available? And another issue that we're seeing is on the liability side where there is real concern about that. If you are somebody who creates works that are being used perhaps to train in AI or that might show up in some of the output from an AI, is that something that you just have to live with? Or is there real liability on the part of the AI platforms for that use? And that's something that is getting a lot of attention and that a lot of clients are thinking about, even folks who may not be directly impacted are looking at these issues and trying to look down the road and see how they might affect them.

SPEAKER_06 13:56

So another issue that I think we've all seen sort of play out in different aspects and some of the reading that we've done on some of these issues is really the role of sort of quote unquote non-humans in creation. So love to hear your thoughts kind of as a gating issue on the question as to whether AI systems, robots, you know, can be credited with the creation of copyright protected works under our current regime of copyright law.

SPEAKER_00 14:23

We have some guidance on that from both the courts and the copyright office itself. The thing that always comes to my mind when I think about this issue is what we kind of call the monkey selfie case, which came up about starting a little over 10 years ago. There was a monkey that took various photos of itself using a camera belonging to wildlife photographer, where it was pressing the button and the camera did what it does, which is make an image. And the photographer David Slater then published a book that contained those images that the monkey had made of itself, monkey selfies using the camera and claimed copyright in the images, although he acknowledged that the monkey was the one who had actually pressed the button on the camera and made the photos. So PETA then filed a lawsuit on behalf of the monkey against Slater and the publisher claiming copyright infringement because the monkey had created the images. And the defendants moved to dismiss for lack of standing under Article III, which is idea there's no case for controversy, and the Copyright Act. The Ninth Circuit ultimately concluded that yes, there was Article III standing, even though PETA didn't have standing to represent the monkey, but the monkey could bring the claim and it would base that on some earlier cases, including one involving dolphins who were affected by some activities of the U.S. Navy. But what it ultimately decided was that the district court, which had dismissed the case, was correct that there was not standing under the Copyright Act because the monkey was not a human and could not create a work therefore that was protectable under the Copyright Act. Relatedly, the Copyright Office separately looked at this issue of who can create a work and what's required. And in 2014, it issued an update of its compendium of copyright office practices and it used the example of a quote, photo taken by a monkey, unquote, as something that it would not register. So the monkey selfie reached the Ninth Circuit Court of Appeals and the Copyright Office Compendium. So if you fast forward now to AI, we've had a few works that folks have tried to register. And one person who's gotten a lot of attention is a guy named Stephen Thaler, who's a computer scientist, and he's tried to get patent protection for AI-related works, and he's also tried to get copyright protection. He created an AI program, which he called the he called Creativity Machine. And he used that program then to create a work that he titled A Recent Visit to Paradise. And the work itself was created by the AI program. He tried to register that. He filed an application claiming that it was a work for hire because he had created Creativity Machine and Creativity Machine had then created a recent visit to paradise. The copyright office refused to register it several times. He requested reconsideration. He had a couple of go-rounds with the copyright office that ultimately rejected it. And he filed a lawsuit in 2022 challenging that decision in the district court in Washington, D.C. Just this past August, the district court granted the copyright office's motion for summary judgment, finding that human authorship was a necessary requirement. And it said that copyright has never been stretched so far as to protect works generated by new forms of technology operating absent any guiding human hand, as plaintiff urges here. And human authorship is a bedrock requirement. And what it did in finding that is look back to some of the earlier copyright cases. You can go back a hundred years to a case called Burroughs Welcome that looked at photography and found that photography is copyrightable because it's not just about the machine, that there is some human standing there pressing the button, deciding which shot to capture, perhaps in a studio, setting up the lighting and the backdrop and maybe arranging poses. But even if they're out on the street, that the photographer ultimately is deciding which moment to capture. And as somebody who likes photographs but is not themselves a good photographer, I can tell you that there is a difference between an image that I make and a difference and an image that is captured by somebody who is actually a good photographer, which would not be any of the ones that I capture. So that's what we've heard from the copyright office and the courts so far on these issues. It really is, though, an interesting question because we don't have a last word on this. We're really sort of starting to look at these issues.

SPEAKER_05 18:58

One of the ones that that intrigues me is translating a piece of original work because it seems to me that that translation ought to be protectable by someone. At least the original author. But apparently, the way it's currently constructed, if that translation is done by a machine, then it is not. And that seems like we're we're something that needs a little more attention.

SPEAKER_00 19:30

It's an interesting question, right? Because you would think of a translation as something that's traditionally done by a human, right? Let's say you have a novel and somebody sits down to translate that in a new edition that is a licensed translation. You typically do have a translator who is sitting there making choices about which word to use and how to structure the language and how best to capture what the original author was trying to convey in the original language. And sometimes it may be a matter of picking different words but capturing the same feeling or sentiment, or trying to decide whether you use a direct translation where the word literally translates, but may not really capture the spirit of the original. So there's a lot that goes into a standard translation standard in the sense of the way we typically think of it being done by a human that is involving a lot of creative choices. But then when you have a translation program that is translating that, I think the copyright office is saying to us, think about that. And if there's not a human involved in that translation, we're not going to register the translation. So now the translation itself is not protectable. And that actually has implications for copyright and for that economic trade-off, because one of the things that the author of a text might be able to do is license it for translation and then sale in another country. But if you can take that original text and translate it using AI and generate a translation that itself is not protectable, is that then copyright infringement? And can the author of the original demand payment for the sale of that translation in another country? So opens up all of these really interesting questions.

SPEAKER_02 21:20

I was going to ask a follow-up to that because the copyright office, after they issued their guidance, held a couple of webinars to explain their position. And they changed it a little bit when they held the webinar. But one of the examples they gave was a translation example, but it was a different kind. They said, imagine a Japanese novelist, she writes a very popular novel in Japan. Her publisher doubts that the book will be sell well in English in the United States, say it's a book about cats, which are very popular in Japan and may not be so popular here. And so if she were to use a translation AI program to do her translation. It's of her work. She's the one who did it. They doubted that it would be accepted for copyright registration by the copyright office in this country. Now, that's not a decision on whether ultimately it's copyright protected, because it's the copyright office, not a court. But a different question since the office said when you register something, you have to point out any expressive elements that it originated with AI. If you're a translator and you're using a translation program to help you make some of the selections you described, Claudia, where would that come out?

SPEAKER_00 22:36

I think that's a really interesting question. If you think about it, what is the translator then doing and how are they using that program? And are they using it to suggest options? And then the human translator is choosing among those options. I think the human in that situation might say to the copyrighted office, well, I'm making the ultimate choice. The ultimate choice is not being made by the AI translation program. But I think it it does ask you to make these very fact-specific case-by-case determinations as to how it's being done and what the extent of the human interaction is with the creation of the new work.

SPEAKER_02 23:16

Thank you.

SPEAKER_06 23:17

So it seems that you know some human creative involvement is required if a work is to enjoy the benefit of copyright protection. Has there been any discussion of how much human involvement is sufficient? How attenuated can human agency be for the product to still have the benefit of this type of protection?

SPEAKER_00 23:39

Yeah, and we're it's something that's still being explored. You know, earlier in 2023, the Copyright Office issued some guidance in February. They issued some guidance that said that a work that combines human-generated text with AI-generated images, for example, could be registered because the human portion of it can be protected, but that the images standing alone couldn't be protected. So if you were to just try to register the images themselves, that would be rejected. But as part of a unitary text where there is human-generated text, it would be protected and registrable to the extent of the text, not the images. We then saw in March of 2023 they issued some guidance on copyrightability in AI and specifically said this is going to have to be a case-by-case inquiry. And we're going to look at whether the contribution from the AI is just a result of what it called mechanical reproduction, or the result of a human author's original mental conception, where the author is ultimately deciding what the form of that work is going to be and what the output will be. We don't know what the bright line is or if a bright line even exists. And one of the things that makes copyright so interesting is you do have this need for case-by-case determination in so many areas. This looks to be another one of those where we may end up with multiple cases addressing this in different contexts where you have enough decisions built up that you get some guidance collectively, but unlikely that any one case is going to be completely determinative of the issue because there are just so many possibilities out there, possible different scenarios where you look at different degrees of human contribution, and some of them may qualify and some of them may not. You know, Thaler has said, look, I created the AI, therefore the program is only making the choices that I've told it to make, and everything that comes out of it is ultimately traceable back to me. The Copyright Office has said no, the DC court has said no, that case is going up on appeal. We'll see what happens with it. But so far, there hasn't been tremendous sympathy for the position, at least from the courts in the copyright office, that you don't have to have a human involved in making the work.

SPEAKER_04 26:01

Along these same lines is a question that comes up a lot in academic circles. The whole idea of fair use. Can you help me understand what that is? Because I've never had a good answer.

SPEAKER_00 26:12

Sure. So fair use is something that comes out of the Copyright Act itself. Section 107 of the Copyright Act says that a what it calls a quote, fair use unquote of a work, which includes things like making copies for purposes such as criticism, comment, news reporting, teaching, scholarship, or research is not an infringement of a copyright in a work. And then lists a number of factors. There are four factors to consider in deciding whether the use is in fact fair. And you know, it's simple to say, well, fair use isn't an infringement. What constitutes a fair use is incredibly fact-specific. And there's a very large body of law out there of cases addressing that issue. If you sat down and tried to look at them and distill a hard and fast rule that you could live by in every situation, you could read every case that's out there. And I don't know that you would necessarily come up with one rule to govern every case that will ever come down in the future. But in general, the statute tells us that the four factors to look at are first the purpose and character of the use, meaning the work that allegedly infringes what is the purpose and character of that? What is the nature of the original copyrighted work? Is it more creative? Is it more factual, for example? The amount and substantiality of the portion used. And in general, courts say, well, we are going to look at that both quantitatively, literally, if it's a text, how many words out of the total number of words? If it's an image, you know, is it all of the image or is it just a piece of the image? And then qualitatively, how important was the amount that was used? And one very famous case involved Gerald Ford's memoirs, where he talked about the pardon that he granted to President Richard Nixon. There was a sterilization in magazines that was officially licensed, and then that was scooped. And in the lawsuit that was brought after that, the court looked at it and said, well, it may have only been a very small amount quantitatively, a few hundred out of maybe thousands of words. But the discussion of why he chose to grant the pardon was extremely important in the history of Ford's life and his public roles that he had. And it's one of the things that people would want to know and why they would want to read the licensed excerpts or buy the book. And therefore, scooping that was quantitatively, perhaps small, but qualitatively very significant. So that's how you look at the amount of substantiality. And then the final factor is the effect of the new work, the allegedly infringing work, on the potential market for or value of the original work. You know, what is the economic impact of this going to be? And when you look at all of this taken together, the concept of fair use is okay, the constitution tells us that to promote progress of the useful arts, we're going to grant this monopoly to people and guarantee them an economic return. And fair use is really saying, look, there are some contexts where we think that that new work has sufficient value to society that we're going to carve out of that monopoly, this particular kind of use. And you look at the things that are in section 107: news reporting, criticism, scholarship, research, all of those are things where you're generating some new work. And what fair use is telling you is to look at that and think about what the value of that new work is. And that's a lot of the discussion that you see in fair use decisions, where courts are grappling with that. And it's not an easy analysis at all. And these are challenging cases for courts. They've been challenging cases for district and appellate courts. The Supreme Court has grappled with fair use a couple of times. And they're always very interesting decisions, I think. And so it's it's not an you know, it's something that where you can say, okay, here are the four factors, these are the things to look at. But then when you actually go to apply it, it is not a simple or straightforward thing in many cases.

SPEAKER_04 30:33

And I guess the emergence of AI and all of its training is just making this a lot more complicated.

unknown 30:41

Yeah.

SPEAKER_04 30:41

Is it okay for the AI to use it?

SPEAKER_00 30:44

That's an interesting question. We don't know. It has not been the subject of a decision yet, but we have a whole bunch of lawsuits that have been filed that may generate a decision and may tell us something about whether what AI is doing in terms of using pre-existing works is a fair use. And it can come up in a couple of different ways. One is the idea that AI programs, particularly large language models, need to ingest works, ingest content in order to be trained to ultimately do what they are being asked to do. And one of the issues that's come up in some of these lawsuits that we've seen filed in the last, call it a year, is do you need to get permission to use pre-existing works for that training purpose? Do you need to get a license? Do you need to pay for that? Or can you go ahead and use it without asking permission and without getting a license and without paying because it's really a fair use and because it's covered by section 107? And there are arguments that are being made on both sides. You see a lot of commentary about it, but we don't have a clear answer from a court yet. So that's one aspect of it that's being litigated. And then another aspect is the output of the AI, where folks are challenging what's being provided then to users who are using the AI and using queries and prompts to generate some kind of result. For example, we've seen Getty images suing stability AI claiming that its images are being used to train the AI and that the output is then reproducing portion of images. One of the claims that Getty made was that it found output where even its watermark was being reproduced from its original images. And you've seen other lawsuits where people have looked at the text that's being generated in response to a prompt. For example, the Universal Music Group sued Anthropic in Tennessee, and in its complaint, did a comparison of text that was generated in response to prompts from users and song lyrics and showed that there was overlap between those two, and in some instances, verbatim copying of the original song lyrics. And so that's being challenged as infringing. And you can imagine that there will be arguments that are now going to be made in those cases that, well, what we're doing and what we're generating really is a fair use. And I'm sure there will be arguments made based on the existing case law, but this is really going to be something new where the courts are looking at this and thinking about it in this new context. And how does the copyright law that we already have apply here?

SPEAKER_04 33:28

Well, you mentioned all the lawsuits. Um isn't the use of copyrighted content to train these models, isn't that what the New York Times lawsuit is against OpenAI?

SPEAKER_00 33:38

It is, and that's one of the most recent ones. And it's it's really interesting. It is an interesting complaint. And if you look at the complaint and also the exhibits, what they've done is try to show through examples that their text from New York Times articles is being used and is showing up in the output from the AI. And you see some of the side-by-side comparisons given in the text of the complaint. And they're challenging both the use as training material, claiming that they believe that their content, which the complaint at least talks about as high-quality content that reflects significant journalistic activity, sometimes investigative journalistic activity that involves research and fact-checking and editorial activity, et cetera, is now being used to train the AI and then is showing up in the output of the AI. And the complaint talks about that as undermining the market that the New York Times is able to exploit for its original works and threatening its economic viability. But then it also points to the output and says, well, if a user can get our output in the AI results and don't need to come to the New York Times website, for example, or to any of the places where we license our content, then that further undercuts our market. And so both of those things are a problem. And you know, you can imagine that the response to that might involve arguments that are based on fair use.

SPEAKER_06 35:16

So digging a little bit more into this concept of fair use. I mean, I think there's probably instances that we can all point to where people are taking a little bit of this, a little bit of that from a copyrighted work to create something that's a little bit new, a little bit more transformative, or just sort of different from the original. So, you know, an example that comes to mind is you know, sampling, where musicians take parts of one composition from another artist to kind of create, let's say, like a bridge to a new song. We've all had those moments when you're listening to like something in the car and you're like, I know this song, and then it takes you back to like something you from like the 1970s or something, right? But those new songs that we're hearing, those are considered original creations. And then on the other hand, right, you know, there's sort of that infamous Supreme Court case where Andy Warhol allegedly infringed on, you know, this a photographer's copyright when he created these sort of silk screen images that were based on a photograph of, you know, the very famous musician Prince. What are sort of the differences there? Like what are the lessons that can be drawn from these very these sort of two different outcomes of taking a little bit of this and a little bit of that and making something new?

SPEAKER_00 36:28

You know, it's interesting, and I think it really gets back to the heart of what is copyright trying to protect and also what is fair use trying to recognize in balancing the author's rights. When you look at music sampling cases, you actually seek claims based on music sampling all the time. And sometimes they're defeated because it's found to be a fair use, sometimes they're not defeated, sometimes those claims were defeated because the court will decide, well, yes, you used something, but it was so small that it was actually de minimis, and therefore it's not copyright infringement. There was a case involving Madonna's song Vogue, where they had sampled a horn hit from a 1976 song, Love Break, and the court decided it was not infringing because it was de minimis, because nobody listening to Vogue would have recognized that horn hit from the original song. And so it essentially was not a use for the purpose of the ordinary observer of the song and therefore couldn't be substantially similar to the original. But you do think about how much is being used, how is it being done? Samples get licensed all the time. But then when you look at the Andy Warhol case, you had a photographer, Lynn Goldsmith, who is well recognized for her portraits. And one of the things that the Supreme Court said in the Warhol v. Goldsmith decision is that the case involved two artists. I think we saw the justices being careful to recognize that there was Warhol, but there was also Goldsmith, and talked a lot about her original work and what the market for it was and how she supported herself as a photographer who's an artist selling her works and licensing her works. And talked about the fact that actually there had been a license to use her work as an artist reference that had been provided to Warhol by publication, Vanity Fair. He then created works for Vanity Fair as illustrations based on her original photograph of Prince, which was taken during photography session that she had with him early in his career. And those works were used. And then years later, after Prince passed away, there was a desire to do a tribute issue. And there was then a licensing from the Andrew Warhol Foundation of one of the images that Warhol had created. And the court looked at that and said, okay, the original was created as an illustration for a magazine article. This new work now is being created and licensed for use as an illustration of a magazine article. And looking at that context, what does Goldsmith do? Well, she licenses her photos as illustrations for magazine articles. And in fact, a number of her photographs of prints had been licensed for other tribute publications after he passed away. And they looked at that and said, okay, the whole of the photograph, or almost the whole of the photograph, was used. It's every expressive element that's in it shows up in the new Warhol work. And it decided that ultimately this is usurping the market for her original. And according to the majority, in that opinion, this was not a fair use. Looking solely at the first factor, purpose and character, the parties had already agreed that the other factors were not at issue. And the one thing for the court to decide was did purpose and character favor Warhol or Goldsmith? And they found that it favored Goldsmith. You had a very strong dissent in that case. And there was actually sort of an interim interesting interplay of the footnotes in the majority and the dissent going back and forth arguing with each other throughout the decision. And there were obviously there's a lot of passion on both sides. There were tremendous number of Amichi on both sides submitting briefs. And, you know, I know from sitting in bar committee meetings where it was discussed that there is a lot of passion just in observers who are not directly involved. You know, if you if the question is, how do you reconcile situations where something is a fair use and something is not? You know, I think that is a question that we grapple with in every single fair use decision that we ever consider. And every time I'm advising a client, whether they're looking at something that's been done with their work and thinking about, is that infringing, or somebody who wants to use something and is worried about what the outcome of that might be and is thinking, do I need to get a license? Could this be considered fair use? You look at all of the independent facts of every particular situation and you make the best judgment that you can. For those who are thoughtful, it's not an easy process. And we've seen a lot of situations where courts have said, yes, this is fair use, no, this is not. I think Warhol is capturing a lot of attention because it's so recent and the Supreme Court hadn't revisited fair use in a while. But I also think that one thing is that I find interesting about the decision is that the majority is careful to say, look, we're just looking at this one factor. And so I'm, you know, there is a discussion of factor four, but the really what was at issue was factor one. And so is it going to be applicable to every single fair use case that comes along? I'm not sure that we know the answer to that, and we may not know for a while.

SPEAKER_02 41:35

Claudia, since you've been talking about photography and copyright, could I take you back to another case involving a photographer, in this case, Associated Press, who took a rather interesting photo of then candidate Barack Obama. And a copyright case arose out of that. Could you discuss that? Because the the Warhol case is very complex. This is a little simpler and may help our audience understand some of the issues because it sort of stands out starkly as not involving some of those complicated issues.

SPEAKER_00 42:10

Well, full disclosure, I was actually involved in that case as counsel for the Associated Press, but it involved an image that was made by Shepard Ferry, who's an artist during Barack Obama's first campaign. Shepard Ferry wanted to support Obama, and so he created a couple of posters that could be put up and indicate support for Barack Obama as a candidate, one of which was an image of President Obama, who was then Senator Obama, with the word hope underneath it. And it was very much of a piece with the style of Shepard Ferry's work. One of the things that he's been quite known for during his career is making very graphically powerful images, often based on pre-existing photographs. And he'll do them sometimes with maybe just two colors or three colors with some text integrated. And that ultimately led to a lawsuit. There was a question initially about what photograph had been used, and there was a lot of sort of speculation online about it. Ultimately, there were some guesses that it had been a photo that had been made by an associated press photographer. There were some discussions, and then Shepard Ferry filed a declaratory judgment lawsuit in the Southern District of New York, asserting that his use of the photograph that he used was a fair use, and therefore his creation of the poster was not infringing. The case ultimately proceeded with both Shepard Ferry and another party as defendants. The other entity that was brought into the lawsuit was basically a merchandising entity that licensed his images and put them on t-shirts and things like that. That and then sold them at various retailer outlets and online. And there were some sort of unexpected complications. At one point, it turned out that Shepard Ferry had fabricated some of the documents that he produced in the case and sought to destroy other documents. He ultimately ended up pleading guilty to obstruction in a criminal case that was brought by the U.S. Attorney's Office in the Southern District of New York and settled out of the case. And that was sort of one piece of it. But then the case proceeded against the merchandising entity. And the district courts ultimately found that it was not a fair use for that entity to use these images and sell t-shirts, and that that kind of commercialization was not a fair use, whatever you might say about the fine art works of Shepard Ferry. We ultimately didn't have a decision involving Shepard Ferry because the piece of the case that involved him settled before that was decided by the court.

SPEAKER_02 44:58

Follow-up to that, OpenAI recently released a system card or assessment of its image generator generative AI program, DAL E, in which it said that because there seemed to be some problems with people asking DAL E to produce an image in the style of certain artists, it was now going to block the program from producing an image in the style of certain artists if they were alive. Could you explain that since from what you said earlier, copyright extends for 70 years beyond the death of the author or the last living author if it's multiple authors? Why have they sort of divided things that way rather than just saying we won't allow it if it's covered by copyright?

unknown 45:46

Yeah.

SPEAKER_00 45:47

You know, I'm I'm not privy to their decision-making process. I don't know why they chose that as the dividing line. It is true, though, that copyright protection does extend beyond the life of any author. For an individual in the U.S., it's the life of the author plus 70 years. It has different durations in different countries, but it is certainly not the case that on the passing of the original human author, copyright protection ends. It can continue for quite a long time. You know, what their thought process was in making that their policy, I don't know. It'll be interesting, I think, with this and all of these other issues involving AI to see what happens. I mean, we've seen, as I said, a number of lawsuits being brought. Some of those have already been subject to initial motion practice based on simply on the pleadings, claiming that the pleadings themselves are defective and that there's been some partial success in knocking out pieces of these cases. Some of them may get past the pleading stage. And then I think we'll need to see what happens with them. We'll need to see if there is a legal challenge to the DALLI approach. You know, it obviously if it's a challenge brought by or on behalf of an artist, well, if you're a current artist who's still alive and you sue, in theory, you may or may not have a claim. If it's an heir, they may or may not have a claim. You may have issues to think about in terms of how that's going to be framed. We know that heirs can have rights, they can inherit copyright interests. Um, we know from copyright termination that even if a grant has been made, heirs can recover those rights and then assert them later on. So I think it'll be interesting to see if somebody does end up bringing a claim based on that. So far, the activity has really been in other areas of AI, but it'll be interesting to see what happens with this notion that AI can reproduce the style of someone and to what extent that style is protectable. You know, you're seeing a little bit of activity in the music space, which may be somewhat informative for this down the road because you've seen various high water marks in terms of copyright lawsuits, the blurred lines case, where there was a jury award of $5 million, finding that the Blurred Lions song by Robin Thick and Pharrell Williams infringed a Marvin Gaye song. Since then, we've seen other lawsuits being brought where the claims are different in the sense that we're not looking at specific notes or specific words. There was a recent case involving Childish Gambino where the claim was that one of his songs sort of mimicked the style of an earlier song, but no claim that notes were the same or that words were the same. There were no bars of music that were being reproduced. And the court was able to get rid of that case relatively quickly. And I'm sure the defendants in the case were pleased with it. Plaintiffs, though, probably not so pleased. And I think you're going to continue to see song cases like that about the style. And it'll be interesting to see whether those cases grappling with the extent to which the style of something is protectable have implications for these other issues in terms of is there something there that can be copied that is protectable by copyright in the United States or in other countries, because it it may vary by jurisdiction depending on exactly what is protected by copyright and each country.

SPEAKER_06 49:26

As we were talking about that Robin Thick song, I had it going back. And then I remember when that case came out, listening to the Marvin Gay version and Robin Thick version, trying to figure out like, oh, how similar are these? How similar are they not?

SPEAKER_00 49:39

So And we've seen those music cases for so many years. You can go back to George Harrison and My Sweet Lord. And I remember being in law school, and my professor bringing in a boombox and playing that and playing He's So Fine and saying, What do you guys think? You're all like, wow, it sounds really similar.

SPEAKER_06 49:56

But you you you were ending your comments there about like different, you mentioned like you know, different legal jurisdictions, right? And how that can play out. And you know, as we've been talking about, as we've been going through this conversation, you know, we'll be talking about Supreme US Supreme Court cases, US copyright law. But, you know, as you, Roland and I well know, and and and and Charles and Mark know from spending too much time maybe with Roland and I talking about the law, is that you know, different jurisdictions can develop and come up with different frameworks for sets of regulations. And sometimes there can be material differences, right, and how a jurisdiction will look at issues and that kind of conflicts of law issue. And we know, you know, the EU in many sort of legal frameworks that they're putting out, including in AI, you know, sort of sometimes set the table for some of these issues. And in the context of the new EU law on AI, you know, your thoughts on how that might play out or how that might guide other jurisdictions, including the US, to kind of consider these issues with respect to copyright rules.

SPEAKER_00 51:01

You know, and it's very new, the new EU law. And so I think people are certainly looking at it and paying attention because it is, you know, very early days, and this is now some guidance, right? You've got the EU saying, okay, this is how we're going to deal with a lot of these AI issues. And I would be surprised if folks in the US on both sides weren't looking at it. You know, how much appetite Congress is going to have to do anything with copyright law in the AI space. I don't know. We've seen relatively recently some changes to copyright law, particularly in the music area. I think, especially going into an election year, we may or may not have an appetite to do more in the immediate near term. But it'll be very interesting to see how the EU law plays out. We've had other areas where the EU law is quite different and it just stays different. And the US doesn't follow a law and says, okay, that works for you guys, but our system is different. We're going to go with our system and we're not going to copy you. And then you have things that really do converge, where you have the EU addressing issues in a way that is seen as instructive and productive for the US system. I do think you're going to see a lot of discussion about it. And I think you're going to see folks watching it very closely to see how it's playing out in actual practice. I think we'll also be interested to see what happens in the UK because the UK is looking at these issues as well and thinking about it. So I think all of that is going to be informative. But I just think it's way too early for us to know. Even if someone could coalesce, some group or coalition could settle on what they think the law ought to be. Are you going to see an appetite for making any changes or an agreement, broadly speaking, that changes are needed, right? Because one thing we have learned from other changes that have been made is you do need multiple parties and multiple interests to come together to actually implement that because it is very difficult to get everybody on board. And there would need to be something in it for lots of different players. So I think it's early days, but it's definitely interesting that that's out there as an example and something for us to watch because it's real world, it's practical. And so we'll be needing to watch it to see what happens.

SPEAKER_06 53:23

I think it's a totally fair point that it is sort of too early to say how this will play out and it's kind of a watch and see mode. I do think that there's one thing that's not too early to say is that the legal profession is not one that moves at lightning speed.

unknown 53:39

Yeah.

SPEAKER_06 53:41

Certainly when you compare it to the pace of sort of technological innovations around AI and the like, right? So with that sort of lens, wait and see more slow-paced legal frameworks developing just as a natural sort of matter of course. You know, how well equipped do you think the current laws and regulations that we have with us here today are positioned to sort of address the issues that are raised by AI?

SPEAKER_00 54:07

You know, it's interesting because we've seen, as we talked at the outset, so many different waves of disruption. Digital technologies are just one example of that. And we've seen so many things change over the years. And nobody can ever perfectly predict what is going to happen and what the new thing is going to be and what we are going to need to grapple with. I do think that copyright is more flexible than it's sometimes given credit for being. And I think because we're in a common law system where courts have the ability to look at the individual facts of cases and grapple with them on a very fact-specific case-by-case basis, that over the years, copyright has shown a fair degree of flexibility, not perfect flexibility. And there has been a need over time for things to be adjusted and things to change. But I wouldn't go into this next phase thinking that copyright law in the US is not capable of addressing AI. I think we do have a framework, we do have tools, we have, you know, as the Supreme Court articulated in the Warhol of Goldsmith case, we have the concept of fair use, we have the concept of derivative works, we have different ways of thinking about it, what infringes and what is protectable by copyright in the first place, not just because of who the author is, but the nature of the work itself. And is every work protectable by copyright? No, not every work is protectable. And that's something the Supreme Court reminded us of. So I think there are more tools in the toolkit than might necessarily appear at first instance. And I wouldn't assume that the copyright law that we have today is not capable of adjusting to this disruption as it has over the years to others.

SPEAKER_06 56:02

That is definitely comforting to hear. Very, very comforting to hear. And so to wrap it up in terms of this conversation, you know, we have a lot of lawyers listening to this podcast, or who will be listening to this podcast in particular with great interest on the insights that you provided. And would love for you to give some thought to our audience about generative AI and how lawyers are using it and some of the potential risks or challenges or pitfalls that they might face as they open themselves up to using this sort of great innovation.

SPEAKER_00 56:38

Well, the one thing I would say is based on the sanctions decisions that have come down recently that we've seen over the summer, and then again within the last month or two, and the news recently coming out of New York where AI-generated search results were submitted to a court and then turned out to be fabricated by the AI to be the product of hallucinations. I think all of us need to be really careful about it, just as we would with any other technology. You know, you can think back to early days of litigation, that you know everything was done on paper, right? And nothing involved computers. And then you had computers, and that was wonderful. And we were able to automate things like document review and do document coding exercises, but you still had to understand what the technology could do and what it couldn't do, what the limitations were, and how things might be missed, for example, in a review project. So I think with AI, it's really important to understand what it's doing when it responds to a query, where that information is coming from, and what you might be able to just accept at face value, and what you really need to check. You know, I I would not expect that anybody I work with is going to be using AI anytime in the near future to do case law research. Just because there have been too already, even with the couple that have issued it, is a couple sanctions decisions, too many. But I, you know, I think there are areas where it's going to be quite useful and quite helpful. And there are going to be parts of legal practice that are made a lot easier and a lot faster and a lot simpler and more efficient and effective because of AI. And the ability to ingest information at large scale and to synthesize it, I think is very valuable. And I think we will come up with creative ways of using it. What I think we'll need to avoid is sort of the blunt instrument approach and assuming that it can do things that it's not really designed for and using it without actually understanding it, which, like every other technology we've ever had, is just not the way to go.

SPEAKER_02 58:51

Well, could you perhaps uh elaborate a little further in this direction? Several courts have now issued local rules in which, if they're precise, they direct them to generative AI. If they're not precise, they speak more broadly about AI. But let's just focus on the generative AI rules where they say either you can't use generative AI to prepare a filing to be submitted to this court, or if you do, you must identify the parts in your submittal that originated with AI. Doesn't that create an extraordinary record-keeping burden for lawyers and for the partners supervising young lawyers who may not be so keen on keeping such records?

SPEAKER_00 59:37

Well, I think one thing lawyers actually are pretty good at is keeping records. So, yes, it does impose a burden, but I think we're going to actually quite rapidly go from a situation where there are just a few of these rules, whether it's because an individual judge in their individual practice rules adopts a rule or because a district as a whole implements rules. And I've been at a number of conferences in the last few months where I've heard judges specifically talk about rules, either that they're implementing or considering implementing. So we're all going to have to pay attention to this. But I think, you know, especially those that focus on generative AI, that is specific enough that most of us should be able to deal with it. I think to the extent it's AI more broadly, that may make us think about where is AI buried in the tools that we already use, right? If we already use a service to do online legal research in case decisions, well, that may be powered by AI. If you get a trademark clearance report, that may be compiled using AI. So I think there's a lot of AI out there that we're already using, they're actually very comfortable with. And I think one way of addressing the court's requirements, which we're obviously going to have to do because they exist, is to think about how we're using AI and be more specific about it and be able to articulate to the court where it might be present in our practice and how and what we think we're doing with it, which is probably not a bad thing for us to think about.

SPEAKER_02 1:01:06

Thank you. I didn't mean to underestimate lawyers' ability to keep records.

SPEAKER_00 1:01:11

We're not perfect, but but we we do it quite a bit.

SPEAKER_04 1:01:17

Claudia, I I must say, if I had professors like you, I think I might have actually finished law school. So again, thank you so much. This is as I expected, this was fascinating. And we really appreciate your taking the time. So uh thank you again very much for joining us.

SPEAKER_00 1:01:37

Thank you for having me. This has been a lot of fun. I love copyright and I love talking about it. So thank you for the opportunity to do that.

SPEAKER_06 1:01:51

We thank the business law section of the American Bar Association for their generous sponsorship of the production of this podcast. We welcome questions and comments from our listeners. Send email to comments at mindthegapdialogues.com. We read all comments and questions and will try to respond in the letters section of a future episode. If you are writing about a particular episode, please do mention the specific episode number. Please do also include pronunciation tips to help us properly say your name when we reply in a subsequent episode. See you next time on Mind the Gap Dialogues on AI.

SPEAKER_03 1:02:53

Thank you for listening to the ABA Business Law Section's podcast series to the extent that the section offers a robust collection of content. To explore more about this topic or to learn about joining the section, visit ambar.org slash bizlaw. That's B I D L A W.