Katherine Forrest: I'm up in Woodstock and, you know, I'm up bright and early recording this, and actually my spouse went out and got me—I'm going to hold up so you can see it, but the audience can't see it—this lovely coffee cup that says on it, “Bread Alone,” which is a coffee shop here in Woodstock. Woodstock is a really fun town. You know, it's where the old music festival. People say to me, “Is it Woodstock, Vermont or Woodstock, New York?” And I didn't even know there was a Woodstock, Vermont. It's Woodstock, New York. And there are seriously a number of people in this town who—I love, I adore this town—but they definitely missed the last bus back in 1969, right? And so the Bread Alone place has my great 2% hot latte, which I have had magically handed to me and I'm gonna drink it right now. And then we're gonna be like jumping into our episode.
Scott Caravello: Well, I have been to Bread Alone. I'm a big fan. I've done actually the last two New Years up in Woodstock, so I'm a huge fan. Love Woodstock Meats and their sandwiches.
Katherine Forrest: Oh yeah, yeah, I don't get sandwiches. I get bones for the dog, yeah.
Scott Caravello: Oh, there you go. One-stop shop.
Katherine Forrest: It's a one-stop shop, and we just ordered our brined turkey this morning. But anyway, so let's just jump in because while this is all fun, people probably are wondering, like, what does this have to do with AI? We don't yet have AI-brined turkeys, but we wanted to talk about some litigation events that have come up and sort of update people on a few things.
Katherine Forrest: And the first thing is the filing of a petition for writ of certiorari to the Supreme Court—which we'll talk about in a moment, what a writ of certiorari to the Supreme Court really is—in a case that concerns, really, copyright rights for AI-generated works. And then second, we had just this week, in terms of this recording week, a decision come out of the UK concerning copyright infringement relating to AI training on image data. And then if we have time, maybe we'll talk about another image generation suit and just sort of talk a little bit about what's been happening in the United States, but those are sort of not events that happened this week. They're just sort of to bring people up to speed to sort of the state of play here.
Scott Caravello: Absolutely. So should I uh maybe just jump in on the case concerning AI-generated artwork? Okay, great.
Katherine Forrest: Let it be known I did not give you a hand motion. I just nodded my head.
Scott Caravello: An affirmative nod, exactly. So that case is Thaler v. Perlmutter. The image at issue is A Recent Entrance to Paradise, which the petitioner, Stephen Thaler, says was autonomously generated by his Creativity Machine, which is an AI system, with no human creative input.
Scott Caravello: And so the fact that there is no human author of the artwork is the defining issue in this case, right? It's a key longstanding principle of copyright law that human authorship of a work is required to receive that copyright protection. When he was seeking registration of A Recent Entrance to Paradise with the Copyright Office, he listed the AI as the sole author and himself as the claimant of the copyright, invoking work-for-hire and ownership-by-property theories.
Katherine Forrest: Right, and, let's sort of just pause here to say that it is not the case that the word author is defined in the statute itself saying that the author has to be human because, of course, when people were drafting the various copyright acts—the 1909, the 1976, many of the amendments since then—the Copyright Office wasn't thinking about whether or not an author was or was not going to be a human. And so, you don't have that kind of immediate clarity in the definition, but there are a variety of ways, and we'll talk about them in a minute, where there's an implication that there's at least a human life form that is associated with the copyright because various dates that you'll talk about, or we'll talk about in just a moment, are sort of triggered off of lifetimes. So when Thaler tried to register the work in the Copyright Office, he had that registration denied twice. And the Copyright Office really stuck to its guidance on this issue, which is AI can be a tool like a camera, but that the Copyright Act is only there to protect the human element, such as the selection, the arrangement, the human touch that is in the artwork. And in 2023, because this case actually goes back a while, the D.C. District Court granted summary judgment to the Copyright Office when Thaler originally challenged the Copyright Office's decision in federal court. And it held that the U.S. copyright laws protect works authored by human beings, not machines. And it leaned on longstanding principles that go back to a case called Burrow-Giles for photographs, and various modern reminders that non-human creators—from gardens that have been shaped by nature or monkeys that have taken selfies—don't qualify for copyright protection. And that work-for-hire—that, that doctrine that Thaler had also used—didn't actually rescue the claim. And, by the way, just as a quick digression on work-for-hire, it basically means that an employer or a commissioning party is considered to be the copyright owner of works created by an employee or a contractor when the creation is within the scope of their employment or engagement. And there's a lot of case law on work-for-hire doctrines, but that's how, for instance, corporations, when they're in the business of creating certain kinds of copyright or copyrightable works, they'll often have their employees sign something and/or it'll be part of their arrangement that the works are works for hire. But notwithstanding that work-for-hire doctrine, if the work isn't copyrightable to begin with, which is what was found here, then there's nothing for the work-for-hire doctrine to capture. So it was sort of a two-step process for Thaler, and his process ended at step one.
Scott Caravello: Yeah, and, you know, on the-on the work-for-hire point, right, because Thaler made the argument that the Copyright Act does contemplate non-human authors because a corporation can be considered the author of a work that's done by an employee. But the D.C. Circuit emphasized the language that I just used, right, that the entity will be considered the author. So it's a fiction versus the human author that has actually created the work and that would otherwise be the author but for this doctrine. And, so—so anyway, the-the D.C. Circuit uh affirmed the district court's ruling and noted the statute's human-centered focus, which we had previewed, right?
Katherine Forrest: Well, but before we get there, I just want to say for our Estonian and Greek audience that I now know we have, I want to just tell you what the D.C. Circuit is. So we have a federal trial court level. Those are the district courts. And then we have the circuit level, which is our appellate level. So you have the D.C. Circuit, then you've got the Second Circuit, the Third Circuit, the Fourth Circuit, the Fifth Circuit, and on. And then ultimately above that, you have the Supreme Court. Now. That was just for a little Estonian-Greek digression. You may continue, Scott.
Scott Caravello: Great call, thank you. But so, under the Copyright Act, right, works created after January 1, 1978, copyright protection lasts for the life of the author plus 70 years. And the statute also uses terms like widow, children, domicile. And so these are phrases that are, in the D.C. Circuit's view, really just referring to humans, right? They don't map onto AI models. You're not talking, at least for now, about an AI model's children. So the appellate panel had also emphasized the Office's longstanding practice of treating computers as tools and not themselves authors. But it's important to note that this is considering the narrow issue where the AI is listed as the sole author, right? The path may remain open for works that are created with a mix of effort. The Copyright Office has released guidelines emphasizing that humans can register AI-assisted outputs if they exercise creative control and aren't naming the model as an author. And the D.C. Circuit's decision doesn't completely shut the door on that possibility either. But it took the view that the Copyright Act does require the work, at least initially, to be created by a natural person.
Katherine Forrest: Yeah, because you can imagine that you've got a prompt where you as a human come up with, I would like, you know, this particular image generation model to create a scene with the following aspects. Uh—It's got rain, it's got X, Y, or Z, it's got, pavement, it doesn't have pavement, you know, it's got a person, the person looks like this, and then the image generation model comes out with something, and then the human then says, “Can you please modify it in the following ways?” And there could be a lot of human input, significant human input in both the ideation and the modification of that, that would then make the human the author. But, back to the Thaler case, you know, there's a key point with it which is really a procedural one too, which is that the whole dispute here is framed by the administrative record that stemmed from the Copyright Office's determination with regard to Thaler's attempt to register the work. And that later statements of Thaler's that he had in fact provided instructions and directions to the generative AI model that created the artwork couldn't be considered because it was sort of frozen in time with the administrative record. And that's an important sort of procedural legal distinction. So Thaler's petition, the petition to the Supreme Court, the cert petition, asks the Supreme Court to consider whether a work can be protected when no human was involved. It does not involve the situation that's more complicated, which is where you've got a record that has this kind of very iterative process that you and I, Scott, were just talking about.
Scott Caravello: Exactly. And so the petition itself, though, is an interesting read because Thaler is building his argument in part for the court to consider that the Copyright Act's purpose, which comes from the Constitution, is designed to, quote, “promote the progress of science and the useful arts by providing this protection.” So in his view, restricting that protection would eliminate critical financial incentives to create and disseminate works using AI, which would effectively eliminate then any incentive to use a tool encouraging and simplifying creation for authors. He also makes the pretty sweeping claim that denying these AI-generated works copyright protection would end copyright protection for many photographs and other works created with technological assistance. Though if I recall correctly, I think the D.C. Circuit had considered some sort of related argument and said that, there-just--that Thaler didn't really provide any support for the argument.
Katherine Forrest: Right, I mean he goes really pretty far in advancing his view that non-human authorship should be acceptable under the Copyright Act and that human authorship should not be required by the Copyright Act and that therefore AI-generated works should fall within its scope. And, you know, in part his view is based upon the fact that corporations have been able to own copyrighted works. And, you know, as we all know, corporations in the United States and most common law countries have been given sort of legal personhood to allow a corporation to own property, to have liabilities, and to have responsibilities. So in Thaler's view, as expressed in the cert petition, the Copyright Office's requirement of human authorship is what he considers extra-statutory policymaking. And--it's interesting. And so, you know, he also invokes a geopolitical sort of argument, which is that the Copyright Office is enforcing a policy that is, as he does say—“deeply hostile” is the phrase, to the use of technology at a time when the United States is seeking to be a world leader in AI. And that's at least a nod to the July 2025 AI Action Plan Executive Order from the White House.
Scott Caravello: Yeah, and it's uh, it’s an interesting signal to the justices, but uh it's gonna be an open question whether the court actually takes the case.
Katherine Forrest: Right, and so now, again for our Estonian and Greek audience—and we, by the way, have other audiences too—but I was so excited that Estonia and Greece, and by the way, my nephew plays professional basketball in Greece, and so I'm- I’m hopeful that, in fact, the entire- his entire basketball team constitutes a significant part of our audience in Greece.
Scott Caravello: There you go.
Katherine Forrest: But maybe not, right? But anyway, let's talk about the moment that is really sort of part of the U.S. court system, and so it may not be familiar to everybody, and even necessarily to some of our tech listeners and people who are not--haven’t really sort of spent a lot of time dwelling in Supreme Court practice. But the position of this Thaler case is at the cert petition portion of the process, right? And what that means is he has made a cert petition, or a petition for certiorari, to the Supreme Court to take the case. So the Supreme Court has not yet even taken the case. So we don't know whether or not the case is fully resolved by the D.C. Circuit's decision or whether the Supreme Court will take it because the Supreme Court doesn't take every case. There are thousands of requests for the Supreme Court to take cases every single year. There are very, very few cases that the Supreme Court must take—for instance, if there's a boundary dispute between two states, Vermont and New Hampshire, about whether or not a river is on one side or the other, or if a river has changed course, whether or not that boundary still is or is not, in fact, the state boundary. That's a particular kind of case that the Supreme Court does take. But most cases, the Supreme Court will accept or not accept based upon a whole variety of complex considerations, but they are presented to the Court in the form of this thing called the writ of certiorari. And it's essentially a request for the Court to take the case. And you have to have four justices on the Supreme Court actually agree to take the case before it's accepted.
Scott Caravello: Yeah, and so there are reasons why the Supreme Court is more likely to hear some kinds of cases and not others. And, for instance, there's what they call a circuit split, which—going back to the concept you explained earlier, Katherine, about our intermediate courts at the federal level—the issue is where two of them in different parts of the country have come to different answers on the same or a similar question. And so the Supreme Court might take it to resolve the issue and create uniformity across the country. And then the other one is if there's a constitutional issue; they consider those cases a lot. And, so moving on, right, if the petition is accepted and decided, it doesn't mean that all the questions of the Copyright Act's applicability to AI-generated works would be fully settled, just the one presented that we've discussed. And it's also worth noting, uh with respect to the dynamics of whether the Court might take the case, is that they previously declined a petition to hear a different case by Thaler on AI inventorship and patent protection. And the question asked was, did the Patent Act categorically restrict the statutory term “inventor” to human beings alone? But that was in 2023.
Katherine Forrest: Right, and there wasn't a circuit split then and there isn't a circuit split on that now. So let's move on to another development in the AI litigation space. And it has to do with a Getty case out of London. And in the London courts, Getty had sued Stability AI in connection with the training of its Stable Diffusion image generator. And there had been some training, at least it was alleged, on Getty's catalog. And the question was whether or not that infringed on copyright. And at trial, Getty dropped their central claim of copyright infringement after acknowledging that the model's training occurred outside of the UK. So they didn't actually get to the ultimate copyright question. And so without proof that the activity of the training took place in Britain, the court couldn't rule on whether or not that training actually violated UK copyright law because the copyright infringement, if any, would have occurred elsewhere. So it narrowed the case to secondary infringement theories and to non-copyright, you know, claims. And, as you know, this whole case, it had the potential to be very interesting because the UK does not have the fair use doctrine in the same way that the United States has. So the conversation about liability and the potential implications for AI in the UK in the context of infringement would really have been significant, interesting, and different than how they might come out in the United States.
Scott Caravello: Oh, absolutely. But on the secondary infringement piece, which the court did still hear, right, Getty also avoided liability there. The High Court in the UK stated that the court's findings were historic and extremely limited in scope, but did reject Getty's secondary copyright infringement claim because, in its words, the AI system does not store or reproduce any copyright works. And so, as a result, Stable Diffusion had not imported a “quote-unquote” infringing copy into the UK as is required for secondary liability because that infringing copy must contain or embody a copyrighted work. And with that finding that Stable Diffusion doesn't store or reproduce any copyrighted work, that condition couldn't be met.
Katherine Forrest: You know, that’s exactly right. And, it’s really, it’s worth a read for those interested in just staying on top of developments in this AI copyright area. The court there in the UK did find a limited trademark infringement tied to Getty-style watermarks that were appearing in generated images. And, but those instances could be very limited. So, with all of that said, the bottom line right now is that with respect to the UK, how training on copyrighted images fits into the country's copyright laws does remain unresolved. So, with that said, I think that's all we've got time for today. We should mention that there are right now pending in federal court in San Francisco, which is the N.D. Cal., a Getty–Stability copyright case, and so there are going to be rulings in the United States where the training occurred on some of that, but it's still pretty much TBD.
Scott Caravello: Exactly, exactly. And so I'm Scott Caravello. Thanks for joining us.
Katherine Forrest: I'm Katherine Forrest. Thanks for joining us.
Scott Caravello: Don't forget to like and subscribe.
