ChatGPT, Anyword, Midjourney. These are among the most popular content generators powered by Artificial Intelligence. But when there’s no human hand behind the creation, what rights does the user of an AI tool have when it comes to copyrighting those sometimes strange images and writings? Well to help answer some of those questions, the U.S. Copyright Office is launching a new initiative to lay out just where its policies stand, and to also hear from the people using generative AI technologies. the Federal Drive with Tom Temin got a chance to speak with Andrew Foglia, deputy director of Policy and International Affairs at the Copyright Office.
Andrew Foglia The beginning is actually a long time ago. The office has actually been following the issue of computer generated works for decades, and even with AI and generative AI, we’ve had events two and three years ago. But the initiative itself is a response to the rapid growth of generative AI technologies. You mentioned ChatGPT, Dolly, things like that. And so it is our effort to look at different issues like the scope of copyright ability in works created using AI technology and the use of copyrighted works to train AI technologies. And it has three major components. The first, we’ve issued new registration guidance to explain how the office will handle works involving AI generated material. Second, we’re going to have public listening sessions to hear from artists, technologists, lawyers involved in these issues about what kind of tools are out there, how they’re using them, and also, what their hopes and concerns are for AI. And then third, we’re going to be soliciting public comments, written public comments to further inform our registration practices and our policy analysis. And we’ve also, by the way, launched a public landing page for AI That’s copyright.gov/AI, which has all the information of all these different parts of the initiative.
Eric White So this still sounds like very much a work in progress. But can you lay out for me what has the office found, so far, regarding copyright issues pertaining to AI created content? Whether it’s a story or a song created by AI, they have all kinds of tools now for that. What has been the general consensus between industry and your office?
Andrew Foglia The general rule at the office is actually a very old rule, it’s not our rule, it’s copyright laws rule. The copyright laws are human authorship requirements, and that’s rooted in the text of both the Constitution and actually the Copyright Act. And the office has written about this before. But because of the new cases were receiving, new applications were receiving involving and generated works, we’ve had to restate that rule to explain it, and we wanted to explain how we’re applying it. So last month we issued a policy statement, which contains a clear statement of the human authorship requirement, its legal roots, and also how we’re applying it to new registration applications. So, for example, it explains that if you are applying to register a work that contains AI generated material, you need to tell us that. You can disclose it as part of your form and we will register parts of your work that have human authorship. You’re not going to give you a registration of the material that was AI generated. If you are, let’s say you already did register a work that generated material. The policy statement says you need to file a supplemental registration. And then if you have questions, it also says you can call us. We have experts who are happy to communicate with you about your registration and make sure that you can properly register your work.
Eric White I promise this is the only specific situation up there. I create a story from one of these AI generated softwares and I say, Hey, I read the story. I want to copyright it, I want to make it my own. And I don’t tell you all that I actually used AI What is the mechanism in place for discovering that? I’m a liar, I guess.
Andrew Foglia So the mechanism to discover you’re a liar are sort of hard to say exactly. There are, the one thing that might happen is if you go to court, so somebody decides to sue you for infringement, or you decide you want to sue somebody for infringing your work. At some point they may discover that you did not in fact write this work. It was AI generated. And that’s going to have a couple of consequences. First, the court can decide to disregard your registration. Second, once the Copyright Office learned, the Copyright Office can cancel your registration. There are also fines from the Copyright Office can levy if you knowingly misstate something on your registration to or other office and it’s material. So basically there are consequences for failing to disclose, and I would be worried about that. And but honestly, we’re trying here to make sure everybody can properly disclose their work. And to the extent that you contributed anything original to that story, you would be able to get a registration, but it would just have to be separate, independent from what’s going on with the AI generated content.
Eric White Got it. So yeah, like you were saying, you’re just going back to the rules still apply. And even though AI may be disruptive for the content creating business. It seems as if, you all are just saying, hey, just stick to the rules and you’ll be ok.
Andrew Foglia With the human authorship requirement it is the same rule. It applies in a way that has been outlined, suggested by past cases. It’s nothing new there, but we appreciate that with the new growth in these technologies. There are a lot of questions and there’s more confusion. And think, it’s important to note that some of these issues, we may not fully anticipate yet. That’s part of the reason we’re having these listening sessions and we’re going to solicit further written comments as we want to learn more about what’s happening. We appreciate that this is a transformative technology, and it’s frankly going to continue changing it. And I don’t think the AI technology is going to stop tomorrow or even whenever we actually solicit comments. But by a hearing from content creators and different industries, we’re going to learn a lot more about what’s going on and where appropriate will issue new guidance.
Eric White You just provided me a perfect segue way into my next question, which is what have you been hearing from different industries coming from these, whether they’re events or just solicited comments that you all got? What have you been hearing from, whether it’s the news creation business or the music creation business?
Andrew Foglia So, frankly, even before we started the initiative, we’d heard from members of the news publishing business. Some curiosity, an interest about how generative A.I. might affect their industry. And other industries, like the movie business, there are more questions about how the registration guidance might apply to, Oh, say, you’re using AI to airbrush out some features of an actor’s face or voice acting, for example. It’s a lot of very case by case specific, like, Hey, I want to use this particular technology to do this thing. Do I have to disclose that?
Eric White Got you. And on that, the people were leaving out of this conversation, so far, are the companies that own the AI algorithms and the actual bots that are creating this content. Where do they stand in all of this? And what rights do they have, if somebody is using their tools to create content to make money for themselves?
Andrew Foglia It’s a very interesting question. As we think about our initiative, we’ve talked about sort of three buckets of big AI and copyright questions. The first, is sort of the authorship question we talked about like when can you register a work that AI Helped to generate. The second, is the training question that I mentioned earlier. When you use copyrighted material to train an AI. And a third set of issues, might do an output infringement, like what happens if you generate output that is somehow infringing, who’s liable under what circumstances? I have to imagine that’s like a it’s going to be a very interesting question. It’s something we are hoping to explore as part of our listening sessions and as part of our later solicitation of public comments. I will say, that we are hoping that technology companies will come to our listening sessions and talk to us. The listening sessions are open to the public, anyone can attend, you can sign up on copyright.gov/AI. And if you want, you can submit a speaker request form and than we’ll be selected speakers to actually be on the panels at those sessions. And again, we are inviting technology companies, we are hoping they participate because we want to learn from them too.
Eric White Yeah, not to mention the confusion of a lot of these AI tools are mostly garnering stuff from past data. What if somebody can identify, hey, my data was the root of this person’s creations and then you have to go back and back and back. Can you go back that far with this AI technology, are you finding?
Andrew Foglia It is a very important question how far you can trace particular data through the learning process. We are hearing, I mean, I think you can even see in some of the lawsuits that are out there. There are different characterizations. If you talk to the plaintiffs in this case versus the, maybe, some of the lawyers representing the technology companies on the other side. And that is an important issue that we are hoping to explore further through our initiative.
Eric White All right. And I just want to touch on the dates that you all have coming up for the public listening sessions. Can you just tell me a little bit about those? I know that you break them down kind of by industries, but can you just lay out what those events entail?
Andrew Foglia I’d be happy to. So the first event is April 19. And that listening session is going to be focused on literary works, including, and this is an odd one for people who are not in the copyright space, but that includes actually software. The second session will be on visual arts, so things involving Dolly, but like not just 2D paintings, but also sculpture, any sort of visual art. The third session will be on audio visual works. Which, maybe movies, some aspects of video games. Video games are an odd case where they cross a lot of lines and copyright world. The fourth will be on music and sound recordings, so it’s one session for each of those kinds of work. And again, we’re inviting artists, technologists, lawyers, academics who work in these areas. Anybody who’s interested in these issues. We are hoping they will listen in, and we are hoping they will register to try to speak at these events, because we are very interested to hear from everyone. I guess, I would just reiterate that we have this new landing page, copyright.gov/AI that has not only links to our past events and policy guidance, it’s going to have speeches and it’s going to have any announcements we make in the future about new policy, about new events, about anything like that. So that’s going to be the best resource for people who are interested in these issues. And I guess one other thing I would mention, I did find this earlier, but if you have questions. Let’s say you read the policy statement or you have a very particular hypothetical that you don’t feel like it’s fully answered by it or you’re confused, call us. We have experts get in touch with us. We have people who are there to help guide you through and make sure that you can do what you want to do and the copyright system.
Eric White OK. And I just want to reiterate that if I do write into you, I’ll be answered by a human right, not a chat bot?
Andrew Foglia That is correct, at least for the time being.