.jpg?cq=50&mw=1180&cpy=Center)
Dive into the evolving world of generative AI with Adaya Esteban as she engages Álvaro Bourkaib, partner at Cuatrecasas, in a riveting discussion. They’ll unpack the legal nuances of generative AI’s impact on professionals and creative works. This episode aims to shed light on how generative AI is reshaping the creative process, authorship, and the IP legal frameworks governing its outputs.
This transcript has been edited for clarity and length.
Zoë Badger: Hello, and welcome to the McKinsey Legal Podcast. I’m your host, Zoë Badger, Associate General Counsel at McKinsey. The McKinsey Legal Podcast is a podcast for legal and business professionals, where we’ll explore legal issues on innovation and creativity that matter to you and the world. This podcast features interviews with global legal and business leaders to learn more about the ways to unlock their innovative and creative capabilities. We’ll highlight stories on innovators in the legal and business world through a lawyer’s perspective, learn from them, and explore how to amplify our own professional trajectory with insights from our guests.
In this episode, my colleague, Adaya Esteban, interviews Álvaro Bourkaib, a partner at Iberian law firm Cuatrecasas, to talk about his legal perspectives on generative AI and, in particular, the implications of generative AI to professionals and to creative works. I hope you will gain a deeper understanding of how generative AI is used in the creative process and the applicable legal frameworks applied to the outputs of a generative AI-assisted creative process. I hope you enjoy this episode.
Adaya Maria Esteban Ruiz: Hello and welcome to the McKinsey Legal Podcast. I’m Adaya Maria Esteban Ruiz, a Senior Contrast Counsel at McKinsey. In today’s episode, we are going to cover gen AI and creative works. AI, more specifically gen AI, is the trendiest topic of the moment, and because of that, I’m excited for this episode and for today’s guest, Álvaro Bourkaib Fernández de Córdoba, a partner at Cuatrecasas, a law firm here in Spain which has been awarded with the Financial Time’s Most Innovative Lawyers Europe 2023 Award.
Welcome and hello, Álvaro. Thank you so much for joining us today. It’s great to have you on our podcast to talk about gen AI and creative works.
Álvaro Bourkaib Fernández de Córdoba: Thank you for having me.
Adaya Maria Esteban Ruiz: Álvaro, before we jump in, I wanted to offer some very interesting background information about you and your career to our listeners. As I have mentioned, you are a partner in Cuatrecasas, specializing in intellectual and industrial property advertising and unfair competition, and focus on online platforms, new technologies, and technology transfer. I have had the privilege to work with you while I was working at Cuatrecasas, so I can confirm that this is quite an impressive career indeed. As you and our listeners can now imagine your connection with the tech world, gen AI and IP rights, makes you the perfect guest for today’s podcast. So, thank you so much for being here.
Álvaro Bourkaib Fernández de Córdoba: Thank you again for having me. Thank you for your kind words and for the invitation. It’s a pleasure to be here.
Gen AI as inspiration for creativity
Adaya Maria Esteban Ruiz: Now that our listeners have a good idea of who we are talking with, I would like you to walk us through this marvelous new world of gen AI that has impacted professionals in every function and in every sector. In particular, we are here to talk about gen AI and its connection with creative works.
We are witnessing an expansion of new applications related to gen AI that produce different types of creative works, whether in the form of images, songs, music, or paintings. Although, maybe “produces” is not the right word here. How do you find this new reality with all these algorithms creating new works for us? Is this threatening? Interesting? What does gen AI mean for creativity, innovation, and technology in general?
Álvaro Bourkaib Fernández de Córdoba: You say, is it threatening? Is it inspiring? I would rather be inclined to say it’s inspiring. Any type of new technology comes with some hurdles and uncertainties, but, if you think from the perspective of how copyright and IP in general have evolved through the years, it’s very linked to the evolution of technology. You can follow a parallel path between the two. I’ll give you a couple of examples. When recordings first came into place, musicians were freaking out. They said, “This is changing the way we see life. This is changing the way we get paid”, because it’s not only limited to live performances. That changed, and thanks to that, we have neighboring rights of performers and musicians.
After that came photographs. Photographs are, if you think of it this way, a machine taking a picture of something that’s outside. It’s purely technical. At first, painters said, “This is not art. This was just a machine doing things”. Now, I’m pretty sure we can all agree that photographs can be a creative work on their own, and not just because they’re made by a machine, they’re detracted from creativity.
I feel the same way, through a different level with gen AI. Gen AI is a very sophisticated and powerful tool, but you can put that tool into the benefit of providing insightful information or data to create something that’s creative by the usual standards. So yes, I see that as an inspiration rather than as the end of our creativity.
The challenges of gen AI and creative works
Adaya Maria Esteban Ruiz: I think this is the answer we all wanted to hear here. I do believe that the photographs example is a great example for what we are witnessing here when it comes to gen AI. We all hear that gen AI also is changing the way we understand creativity and innovation, as you already mentioned. Doing many things, augmenting human creativity, expanding creative possibilities, music, audio content, surely in the future, many other activities as well. But we also read about the risks of this technology, such as hallucinations, discrimination, lack of accuracy, sometimes lack of efficiency due to lack of development. Which are the top three most relevant legal challenges you believe we will be facing when it comes to Gen AI and creative works?
Álvaro Bourkaib Fernández de Córdoba: This is a very good question. I always refer to three or four. I will stick to three. One of them is the exhaustion of alternatives and let me explain that. If you think of a 12-bar blues, a 12-bar blues is a very known structure of music. It’s based on a very specific harmony. So, if you think of it this way, there’s just a limited number of combinations through which you can drive a 12-bar blues.
If you put that into a machine, a machine could make all the theoretically possible combinations of that, which could exhaust the alternatives that a human being could have to be creative. This is a very small example that I made, but if you expand that to other types of creative works, the power that AI has within its boundaries, it can just provide you with all the alternatives, and that can exhaust the potential creativity of humans. And we’ll discuss afterward whether or not this can be protected, but from a human-centric perspective, if the machine can do all the alternatives, what’s left to you? So, that would be the first. Just to be clear, I don’t think that’s a stopper for creativity, but this is one of the challenges I can see in using this technology.
Second is what I call decoding creativity. I’m sure you’ve heard the discussions around creativity in machine-produced outcomes sometimes boil down to whether there’s a mathematical rule underlying the creation itself. And if you, compare that to music to pick that as an example, music is fundamentally a mathematical relationship between different chords and music and timbres. If you reduce creativity and works of authorship to mathematical rules to argue whether or not something is creative when it comes out of gen AI, you are putting a protocol in the creative process. I think that is a very complex way of thinking because creativity is a complex process in the human mind, and if you translate that into little steps, small bits of information that can have a rule or some sort of relationship between each other, which is what the machine does, then we’re defining creativity in a very strange way. So that would be the second.
The third is maybe more attached to the legal environment, but the third is paternity. How far can you push a gen AI to produce a “creative result”, and what part of that comes from a human being as opposed to the machine? If that question leads us to who is the author of the result, you get in a very difficult situation when it comes to attributing paternity to the result. In a nutshell, that would be the three challenges.
Authorship in humans or machines
Adaya Maria Esteban Ruiz: I totally agree with those. I haven’t thought about the first one before, but I also see a space for improvement when it comes to other opportunities, and other combinations as well. And with regards to the last one, I think it’s a good one because we can also link this one of our questions, which is, how can we protect the result created by the gen AI? How is gen AI currently regulated? How do you envision this protection is going to be articulated?
Álvaro Bourkaib Fernández de Córdoba: Well, now the general rule of course, I would say for all countries, is that IP is very human-centric. Authorship or inventorship is only recognized to belong to human beings. This is partially because the rules that govern IP are old, but I’m not saying they should be updated. They’re thought as an attribution of rights to human beings, thanks to the contribution one is making to the whole society through creative words or inventions.
It has been tested in court in several countries, whether or not an AI can be considered to be an author or an inventor. Everywhere has been dismissed partially because the rules and the law says that it has to be a human being, but also because this is not the way this is conceived. There is at least one notable exception, which is that in Australia there was a court that accepted an AI as an inventor. That court has had some insightful takes, where that court said, basically, that the concept of inventor or author, well maybe the author is a different one, but the inventor is an agent, just like a dishwasher doesn’t mean that it’s a person washing the dishes, it can be a machine doing that job. So, inventor to this court meant whatever produces the invention. This was overturned by the Superior Court on appeal, but that reasoning gives a slight hint of what’s coming up next in the future.
And moving to that, I see the future in the short run as imposing rules on input. Let me explain that. If you set the boundaries or the thresholds on how creative the input needs to be, so that you sufficiently influence the machine to produce a certain result, so that the result can be attributed to the input in the first place, this is something that we will be discussing in the next years: This connection between the input and the output. We know for sure that outputs are not controlled. For many types of AI, you do not control what the output is, but, if you think of a creative input that may lead to a creative output, then if we can link those two, you could argue that there’s creativity in this process. Just like in the example of the photograph, the machine does all the work, but the selection of the space, what is being photographed, the moment, the lighting, whatever you’re deciding creatively on the photograph, is what defines the work of authorship on its own.
It’s basically the same type of reasoning, but applying a different technology, which is far more complicated for our human minds, but still producing the same creative result. This is in the short term how I see the discussion on ownership and creativity using AI.
In the long run, let’s be a little bit sci-fi here, but not too much. Why not think of smarter, if I can use that word, smarter AIs as some sort of legal person. I’ll explain that. This is not unusual in a legal environment. If you think of companies; companies are legal persons. They’re legal entities and they have rights, and they have obligations on their own. This is just something that we as a society and, particularly, we as lawyers, decided we invented just because it’s convenient.
It’s convenient to allocate rights and liabilities on a specific container, which is a company. Why not think of that for AIs too? I’m not saying with all the rights that a human being has. Definitely not. But why not actually give them some sort of rights and obligations in the future? And it’s not just me saying that. The European Commission, three or four years ago, issued a white paper on the civil law implications of AI and one of the things they consider is whether they should start thinking of attributing some sort of personality to AIs as a subject of rights and obligations.
So maybe, if we think this way, the person who legally owns the AI, and controls the AI, can be the ultimate beneficiary of what the AI does, just like the shareholders in a company are the ultimate beneficiaries of the assets of the company. And just a warning, because we’re talking about, ownership of the result of an AI, remember to look at the terms of service. Because some of them, not all of them, but some of them address this issue. If we agree that there are some authorship rights derived from the output of an AI, then we might be contractually bound by what we signed in the terms of use and that may belong to the user or not. So, read that through, because this is already regulated in some of the terms.
The fair use doctrine and training AI models
Adaya Maria Esteban Ruiz: Maybe even the terms of service are going faster than the regulation up to now, up to today at least. So, be careful!
Maybe a couple of lines with regard to the last one, because of the importance and I think our listeners would like to know a little bit more about the fair use doctrine and how this will be or could be applied to training datasets when it comes to gen AI. Please help us through this doctrine a little bit.
Álvaro Bourkaib Fernández de Córdoba: Yes, this is being very debated out of court and in court. It’s not the first time that using a technology is put in front of a court under the fair use doctrine. I would say the landmark decision, so far, is the Authors Guild versus Google Books case in the Second Circuit.
There it was decided whether the scanned reproduction of a large amount of books made by Google for the Google Books service was an infringement of copyright rights on their underlying books or if it was covered by the fair use doctrine. This was a complex and long case, but it all boils down to one of the four factors in the fair use analysis, which is whether or not the use is transformative. What the court says is that even though they were using the whole book, and they were scanning and reproducing the whole book electronically, which was disfavored in terms of fair use, the use they were making was limited to small snippets. More importantly, it was transformative in nature because it was providing a different use to the book. It was not reading the book as a usual reader would do. It was providing the information that was within the book, thanks to the technology. So, it was converting the contents, and information, and that specific use of the books was transformative in nature. If you think of that now, there’s a plausible argument that using AI to use information protected by copyright to train an AI in a way that the AI can produce an output afterwards, and so the output would not be reproducing the input, but providing something else, just like gen AI.
Gen AI would use the mathematical links between the words, if you think of the book, to provide a newly generated response. So, the question is whether scrolling through all that information, which is protected by copyright, is a copyright infringement on its own? We’ll get to see, but not in the distant future. There is, at least that I know of, two court cases in the U.S., two class actions. One brought by a group of visual artists against Stability AI (for Stable Diffusion), and another one was filed in September last year by the Authors Guild against OpenAI. So, we’ll eventually see what the court says on the fair use doctrine, which by the way is not applicable across the globe. We have different rules in Europe, as you know very well, we have a limited number of exceptions and limitations to corporate rights, but still, we have a promising future ahead of us because there are limitations that, depending on how you interpret them, may allow this sort of training for the AI.
Adaya Maria Esteban Ruiz: I will be happy to hear about how this is sorted out in the future. I’m guessing our listeners will as well. I think we are at the end of our podcast but thank you so much. This has been a really interesting conversation, and it’s been a pleasure to have you here.
Álvaro Bourkaib Fernández de Córdoba: It’s been my pleasure to be here.
Zoë Badger: Thank you for joining us on this episode of the McKinsey Legal Podcast on generative AI and creative works. Join us for the next episode where we explore legal issues on innovation and creativity that matter to you and the world. This episode is a production of the McKinsey Legal Department, was produced by Stephanie Spangler, co-hosted and written by Adaya Esteban, and edited by Carmen Silva. The original music is by David Shaporov.