Generative AI and Copyright: Collision is Inevitable

Episode Notes

Transcript

[Music]

Intro: Welcome to The Digital Edge with Sharon Nelson and Jim Calloway. Your hosts, both legal technologists, authors, and lecturers, invite industry professionals to discuss a new topic related to lawyers and technology. You are listening to Legal Talk Network.

Sharon D. Nelson: Welcome to the 184^th Edition of The Digital Edge Lawyers and Technology. We’re glad to have you with us. I’m Sharon Nelson, President of Sensei Enterprises, an information technology, cybersecurity and digital forensics firm in Fairfax, Virginia.

Jim Calloway: And I’m Jim Calloway, Director of the Oklahoma Bar Associations Management Assistance Program. Today, our topic is Generative AI and Copyright: Collision is Inevitable. Our guest today is Vedia Jones-Richardson, a Partner heading the Transactional Practice Group at Olive and Olive, an intellectual property law firm in Durham, North Carolina.

Her practice focuses on consumer products, computer, communications, educational arts, and entertainment clients. She has served in the American Bar Association in many roles as Chair of the Law Practice Division on various ABA Commissions for Technology and Client Development Issues, as an elected fellow of the American Bar Foundation and has also been a fellow in the College of Law Practice Management. A Founding Board Member of InternetBar.Org, and has served on the boards of many arts and cultural organizations. Thanks for joining us today Vedia.

Vedia Jones-Richardson: I’m happy to be with you.

Sharon D. Nelson: Well, Vedia, most people I think don’t understand the connection between generative AI and copyright law. Can you help them understand what that connection is?

Vedia Jones-Richardson: Well, I would say that the connection is becoming realized as people recognize that generative AI tools are becoming able to produce the kinds of material that typically have been protected by copyright law. Textual and visual and multimedia material even software code, which essentially it’s just a form of textual material all can be covered by copyright claims.

Jim Calloway: The next big question everyone is asking is whether the AI outputs, whether its images or text can be afforded copyright protection, what would you tell them?

Vedia Jones-Richardson: Well, the simple answer is no. The courts and the copyright office have made it clear that there will be no copyright coverage for output created solely by generative AI tools.

Sharon D. Nelson: But some have opined that there could be some copyright protection if the work was partially created by AI and partially created by human being, so where do you think they’re going to go with that?

Vedia Jones-Richardson: Well, that’s true but to a certain degree. If a human contributes additional creative material to something that was produced by AI, then the additional material created by the human can be covered by copyright.  So what that means is essentially is that if you use AI to produce material that you want to claim the copyright for, you can use AI as a starting place but then you need to be prepared to make the material your own by actually adding your own creative output.

Nevertheless, AI can be a good way to get the creative juices flowing. But what the AI produces itself is not yours.

Jim Calloway: The Supreme Court has held since the 1884 case, Burrow-Giles Lithographic Company v. Sarony that photographs can be entitled to copyright protection or the photographer makes decisions regarding creative elements, such as composition, arrangement and lighting. Generative AI programs might be seen as another tool like a camera that can be used by human authors to create copyrighted works. Well what do you think of that legal theory?

Vedia Jones-Richardson: I think the situation is slightly different in the case of AI as compared to photography, at least in most cases. A photographer is going to make a lot of creative choices about the subject of the photograph and the capabilities of that tool, the camera when taking a photograph.

And in the case you mentioned, the court actually talked about all the things that the photographer did to set up the scene in advance, to frame the image, to set up the camera, the lighting, even they talked about the way the folds in the fabric were placed so that they could produce a pleasing composition. And then the photographer did use the camera as a tool, but it’s a tool kind of like the way we might use a word processor or a computer graphics program to produce material.

(00:05:04)

We have a lot more control over that tool than we usually would have with an AI tool. So generally speaking, we do not exercise sufficient control over the outcome of the AI tool so that we can think we should actually claim credit for that output.

Jim Calloway: Before we move on to our next segment, let’s take a quick commercial break.

[Music]

Advertiser: Built for lawyers, Nota’s cloud-based business banking is perfect for your solo or small law firm. Easily set up categories to track income and expenses in a way that integrates smoothly with your practice management, accounting, and payment systems. Create instant transfers to your IOLTA that can pass any audit, print checks from your desk, and get excellent service from reps who only work with attorneys online at trustnota.com/legal. Nota, banking built for law firms like yours. Terms and conditions may apply.

[Music]

Jud Pierce: Workers Comp Matters is a podcast dedicated to exploring the laws, the landmark cases and the true stories that define our workers’ compensation system. I’m Jud Pierce and together with Alan Pierce, we host a different guest each month as we bring to life this diverse area of the law. Join us on Workers Comp Matters on the Legal Talk Network.

[Music]

Sharon D. Nelson: Welcome back to The Digital Edge on the Legal Talk Network. Today, our subject is Generative AI and Copyright: Collision is Inevitable. Our guest today is Vedia Jones-Richardson, a partner heading the Transactional Practice Group at Olive and Olive, an intellectual property law firm in Durham, North Carolina.

Her practice focuses on consumer products, computers, communications, educational, arts and entertainment clients. Vedia, one law professor suggested that a human user who enters a text prompt into an AI program may be asking Dolly to produce a painting of hedgehogs, having a tea party on the beach, which sounds quite charming to me, they suggested that was just contributing nothing more than idea to the finished work.

So according to that argument, the output image lacks a human author and cannot be copyrighted. What are your thoughts on that?

Vedia Jones-Richardson: Well, I think that’s essentially correct. So first of all, when we’re talking about what this professor is talking about, we’re talking about the text prompt which is the input, not the output. Up until now we’ve been talking essentially about protection for the output but in this case what we’re talking about is what is said to the system in order to get it to produce the output.

So we can prompt and coach an AI tool to produce the desired output but we don’t actually control that output. We tell it what we’re thinking about and then we let it go and give us its material, its thoughts, if you will. Again, we’re not really doing much more with respect to that output then prompting the computer to produce it. And I think another important consideration to think about here is that copyright doesn’t cover ideas. It only covers the way someone expresses or executes on an idea.

So if I express an idea or a concept, the way I put it, the words I choose and even the way I frame the idea, those are my expression or execution of the idea, how I’ve decided to express that idea, and that’s something I can claim a copyright for, but someone else might discuss the same idea in different words, which would be a different expression with a different copyright, their own material.

So that can be some additional way to think about what happens with AI product because what it then gives you after the prompt is the expression of the AI, not yours and then the human must go back in and edit that or change it in some way and there are some additional elements at that point that copyright law does consider protectable under copyright protection.

Things like the way you choose what gets kept or the way you organize the concepts in order to present them in the most understandable way. So there’s a concept in copyright law that calls selection and organization.

(00:10:02)

Even if you’re taking facts and putting them in a certain order in order to help the idea be more understandable that can be the nature of input that allows the human to claim that he added something to the output. Generally speaking, without that additional human expression, you don’t really have anything other than what came from the AI.

Sharon D. Nelson: I kind of get that but I was so intrigued by the painting of the hedgehogs having a tea party on the beach. I thought to myself well what if you got that? You know, it’s not copyrightable but I removed the beach part and put a forest in. I can copyright that, right?

Vedia Jones-Richardson: Well, it depends, it depends. I mean did you paint that forest? Did you take it from a photograph that was actually somebody else’s creative work?

Sharon D. Nelson: I’m an artist. I did the forest, it’s my forest.

Vedia Jones-Richardson: Okay so then you have added something that is in itself copyrightable and then maybe even the whole of the composition, maybe copyrightable but you must still extract the hedgehog image that the AI created from the overall piece as something that is not copyrightable.

Sharon D. Nelson: I’m still going to go and have fun with that. All right, your turn Jim.

Jim Calloway: You know, I’m not sure many lawyers have thought of this yet, but if a lawyer submits a brief or other document to a court and it turns out to contain copyrighted material, can the lawyer be sanctioned by the court or sued by the copyright holder?

Vedia Jones-Richardson: Well, I think lots of lawyers have been thinking about ways to use AI and they’re working with varying results. In particular, the issue you put forth is whether the material would be — if it’s infringing, whether that would be a problem and we really haven’t been talking that much yet about infringement. We’ve been talking more about which aspects of an AI generated product are protectable.

But if you have gotten material from elsewhere and it’s protectable or copyright protected material, and then you’re using it in the output, then you have to question whether the original material can still be infringed and it certainly can. So that doesn’t change simply because it’s produced using AI, so whenever you’re using any material that you’ve not created yourself, you should always be concerned about whether you have a right to use it.

It’s not always easy to determine whether you do have that right and it’s even harder when the material is coming from someplace, you don’t even know because the AI hasn’t told you. So you could wind up infringing without even knowing it when using material produced by AI.

Sharon D. Nelson: Does it matter Vedia whether the lawyer knew or did not know that the material submitted contained copyrighted material of someone else, does that matter?

Vedia Jones-Richardson: Not necessarily and that therein lies the danger because a person can be held liable for infringement without even realizing that the material that was being used was infringing material. Of course, that can be a mitigating factor. Not every infringement is going to result in the worst treatment. But this is definitely a risk.

So I think though that the bigger issue in the case of a lawyer using someone else’s material is not really infringement. I’m not sure if you’re familiar with the recent case, I think it was a New York situation although the lawyers were from New York and New Jersey, where they submitted a brief to the court that contained cases that did not exist.

So any time a lawyer is using someone else’s material or material that he or she did not themselves create, they’ve got that additional obligation to make sure that the material is legally reliable.

Sharon D. Nelson: Well, we’re still waiting to see what the judge says about that. My suspicion is sanctions are forthcoming, but we’ll see if I’m right.

Jim Calloway: Vedia, I did notice in conjunction with that that I think the third judge this week has issued an order saying you have to do an AI disclosure and say you’ve read all the cases that have brought up and that’s going to add lots of litigation cost, isn’t it?

Vedia Jones-Richardson: Well yeah, people are coming up with all kinds of ways to work through this situation and to help people understand what’s coming from where.

(00:15:02)

And the question is not really whether the material comes from AI but is this material reliable? Did you really look at it and decide for yourself that you agree with it.

Jim Calloway: Read the cases you cite.

Vedia Jones-Richardson: Absolutely.

Jim Calloway: Does the AI training process infringe copyrights and other types of works?

Vedia Jones-Richardson: The training process and you’re probably going to need to help me here because I’m not entirely sure that I understand it. To me, it’s a process of gathering information from many different sources and using that to train the system as to how to produce the output, whether it’s from a connotative standpoint or a stylistic standpoint but a lot of material that will be in various places is fed into the system.

And then the question is when that material goes in, is that in itself an infringement of that material, and that’s still a question being decided. Right now, the common consensus seems to be or the official view is that it’s an example of fair use. And I can see that view in a way because to a certain extent, we’re looking at going out and seeing material, reading material, much in the way a human might do that when a human is training himself.

When a kid learns language, he’s listening to a lot of different words spoken and understanding the responses and using that to train his brain as to what the meaning of those words is; and likewise, when we do research in a library we are looking at a lot of different pieces of work and we’re putting that into our brain to train ourselves how to perceive that subject or topic that we are researching.

And all of that input is then melded into what we want to think of it and to produce as our output. But then that output is often quite different from the form in which it went in. So what copyright law looks at is did you copy? Whether what you took in was then used to inform yourself and produce an entirely different output or whether what was put in was spit right back out and even if it does spit out in combination with other things that were put in and made some form of compilation or collection of things, that’s not transforming that thing that went in into something new, that is yours alone.

If that output reflects the input directly then that could be considered copying but training, as I understand it, is a slightly different process. So for me, the question is whether the source material is providing training in terms of expanding our knowledge, background knowledge, or whether it’s being reused for its own sake and then the output is the same product as the input even if it’s combined with other inputs to create a collective output.

Sharon D. Nelson: It is interesting. And I will add that, as far as I know, the AI training process, just like the rest of it, is a black box and so long as we can’t see what’s happening, we can’t answer the question.

Vedia Jones-Richardson: Well, there have been a lot of cases out there where artists and authors are trying to get together with groups that can help represent them to bring claims against the creators of these AI tools and ask that they be paid for having provided the base knowledge that forms the AI’s collective knowledge.

Sharon D. Nelson: That will be fascinating to watch, that’s for sure.

Jim Calloway: Before we move on to our next segment, let’s take a quick commercial break.

[Music]

Advertiser: The ABA Journal Legal Rebels podcast features the men and women in the legal profession who aren’t satisfied with good enough. These are the people who are changing the way law is practiced and setting the standards that will define the profession in the future. Each episode, we share their story. To hear insights from those with an eye fixed towards tomorrow, follow the Legal Rebels podcast, part of the Legal Talk Network.

[Music]

(00:19:55)

J. Craig Williams: Today’s legal news is rarely as straightforward as the headlines that accompany them. On Lawyer 2 Lawyer, we provide the legal perspective you need to better understand the current events that shape our society. Join me Craig Williams and a wide variety of industry experts as we break down the top stories. Follow Lawyer 2 Lawyer on the Legal Talk Network or wherever you subscribe to podcasts.

[Music]

Sharon D. Nelson: Welcome back to The Digital Edge on the Legal Talk Network. Today our subject is Generative AI and Copyright: Collision is Inevitable. Our guest today is Vedia Jones-Richardson, a partner heading the Transactional Practice Group at Olive and Olive, an intellectual property law firm in Durham, North Carolina.

Her practice focuses on consumer products, computer, communications, educational, arts and entertainment clients. Vedia, fair use is a big issue at the moment with the United States Supreme Court’s recent decision on Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith, a case that considered whether a series of images Andy Warhol created of prints were adequately transformative under the Fair Use Doctrine of the photograph Warhol used for reference.

What do you think of the outcome and how does that impact issues related to AI?

Vedia Jones-Richardson: Well, I am not sure that that case really provides us with a lot of guidance. In my opinion, that case was mostly about using something beyond the permitted purpose and then trying to say that that was allowed because it was different from the permitted purpose.

But if you needed permission in the first place, my question is then why didn’t you need permission for the second place. It does have some relevance to AI generated material on that issue of what is transformative and therefore not considered infringing, which some people say does apply to AI output. But even the Warhol case did not resolve that question.

Jim Calloway: Vedia, I think we touched on this earlier but in a guidance document recently released by the US Copyright Office, the office clarified its stance on AI generated works and their eligibility for copyright protection. It’s probably a good way to conclude this podcast by saying can you summarize for our listeners what that stance is?

Vedia Jones-Richardson: The Copyright Office essentially clarified that it will not offer copyright protection to the material generated by AI unless some additional substantive human created material is added to the AI generated material. It will not protect the work. So if you’re applying to the Copyright Office to get a registration for work that you believe is copyrightable, you basically have to specify whether AI was used in creating that working.

Copyright Office’s changing its application to make sure there’s a place to indicate that and in fact, I think it’s a about a week from now they’re conducting a major training webinar on how to prepare applications that involve some degree of AI generated material in them but essentially the way they’re looking at it is that you must specify what was AI generated and what’s not and just as in the case of any copyright registration application where you have the material that was yours and some material that was from elsewhere, you have to disclaim that material that’s not yours as not being covered by the copyright and then specify which part you are claiming as yours.

Sharon D. Nelson: Well, we want to thank you for joining us today. This is a complex subject. It’s certainly a very current subject and there’s lots of arguments about it and I don’t know how we’re ever going to figure out what the Black Box implications are but I can see that those cases are going to come. So it was kind of nice to have a chance to go over the copyright implications because I know you’re probably reading just as Jim and I are, you’re probably reading an article or two every day about something that’s happening in this area. So it was a great gift.

We were able to give this particular podcast today on this topic, which is so hot. So thank you very much for being with us.

Vedia Jones-Richardson: Well absolutely it was my pleasure to talk with you about it.

Sharon D. Nelson: That does it for this edition of The Digital Edge Lawyers and Technology. And remember you can subscribe to all the editions of this podcast at legaltalknetwork.com or an Apple podcast. And if you enjoyed our podcast, please rate us in Apple podcasts.

Jim Calloway: Thanks for joining us. Good bye Ms. Sharon.

(00:25:00)

Sharon D. Nelson: Happy trails, cowboy.

[Music]