Introduction
It is crystal-clear that works generated through artificial intelligence (AI) systems are starting to compete with works made exclusively by humans, with or without the help of technology. Advances in the field of AI are moving the state of the art almost month by month towards limits that were unimaginable just a few years ago, thanks in part to the powerful investment that this area of the technology industry is receiving from some of the most powerful companies on the planet. And all this technological disruption is also affecting, either directly or indirectly, almost any field of art, culture and entertainment.
Indirectly, you only need to look at any content recommendation system. There is a lot of content, more and more, and also more types of audiences. The problem of matching content with audience has long been solved by AI systems, resulting in companies like Netflix1 evolving from being mere distributors and creators of content to being technology companies themselves. The line between one type of company and the other is becoming increasingly blurred, with capitalism working its magic on those companies that strike a balance between the purely technological and creative spheres. The impact that the use of this technology in the distribution of content has on its creation, we will leave for another day, because in one way or another, it evidently has it.
Directly, and beyond the already classic example of The Next Rembrandt, I would like to talk about a new category of artists per se, the so-called AI artists2. One of its greatest exponents is Mario Klingemann, an artist who has been using AI to generate works for 4 or 5 years now. Klingemann approaches the artistic process in a different way, curating what the AI generates, and this is not a trivial matter, as this will probably end up being the trend in the generation of most of the works referred to in Article 2 of the Berne Convention. This is a point that is worth dwelling on, because at the end of the day what is happening is that the distance between the idea itself and its materialization is shortening.
Towards the materialization of the 'idea'. The current impact of AI on the human creative process
Human beings have been trying to find tools to give shape to our ideas since the ancestral beginnings of mankind. Progressively we have been creating technology with which to shape our ideas, from mere utensils, such as a paintbrush or a pen, to complex computer programs, such as Adobe Photoshop, and finally, in recent years, AI systems with which to materialize what we have in our heads through simple prompts3.
In other words, as a human species what we have been doing is progressively adding layers of abstraction to the artistic and creative process, thus separating technical capabilities from purely creative ones4. As an example of this, while in the past, in order to create a work of art (e.g. oil painting a realistic painting) required great technical expertise (as well as creativity and, in terms of copyright, originality), nowadays, as long as you know how to use a certain software, you can create works with almost none of the technical skills required to create the same works using traditional tools. In this way, what is truly relevant is creativity and originality, leaving technical expertise behind.
If we transfer all this to the creative process, we have as a result a progressive disappearance of the requirement that the generation of the idea and its expression converge in the same subject. In the past, in order to generate a work, it was not only necessary to have an idea, but also to know how to execute it, requiring, in most cases, a certain technical expertise. However, with the advance of technology, we are getting closer and closer to being able to materialize such an idea without having to practically touch the work (in fact, this seems to be the direction we are heading in, especially if we look at the progress in the field of biotechnology and neurotechnology5). Let's see it more graphically in the following figure6:
Through this diagram, the authors of the aforementioned article describe the creative process in three phases (conception, execution and writing), emphasizing the importance of iterations throughout the process to reach the desired work.
- The 'conception', as its name suggests, would have a human origin, insofar as it is a product of our creativity, and includes all those design choices relating to the substance of the work (see the choice of genre, style, technique used, materials emulated, etc.).
- On the other hand, 'execution' would be the material production of the work. Traditionally, human intervention has been crucial in this phase (think of painters or sculptors), but as technological inventions of various kinds have appeared, machines have progressively acquired an increasingly important role in the execution, going from being mere accessory tools to the execution process (see a camera in the case of photographs, or recording devices in the case of music), to being the actual executors of the raw versions of the work.
- And finally, in the phase that the authors call 'revision', certain creative choices are made to give the work its final touch, making it ready, for example, to be commercially exploited.
In this way, and to illustrate as simply as possible what I am referring to, the above figure would look as follows7:
We are witnessing an unprecedented rise to prominence of machines (and to be precise, AI systems) in the execution phase of the creative process.
But... in practice, what is happening?
The consequence of all of the above is an acceleration of the creative process, making it more practical, accessible and therefore more attractive to creators and artists. In fact, some publishing groups I have had the opportunity to work with are already assuming (and making it clear in contracts) that the artists they work with use generative AI tools such as Midjourney or Stable Diffusion to create the illustrations they request, and the same happens with texts in the case of literary works and tools such as Chat-GPT. Let's look at a current example of what is happening.
Recently, while browsing through the Barcelona Bar Association magazine 'Món Jurídic', the April/May 2023 edition8, I came across the following advertisement for Barcelona’s Joyeria Grau:
As evidenced in the upper right corner of the ad, the image was created with AI. We don't know the reasons that led the beneficiaries of the advertisement to opt for an AI-created model versus a real one. Perhaps for a reduction in cost, perhaps to embrace innovation. But from the style of the image, it does not seem unreasonable to think that its raw execution has been the product of 'prompts' that a creative has supplied to an AI system, to later give it its final touch, leaving it ready to be used as the back cover of the magazine. What probably evidences the above, is the evolution of the creative process in the sense that I have expressed in figure nº 2.
With all this, perhaps soon there will no longer be only a few artists and creatives who use AI to generate art -as was the case with Klingemann a few years ago-, and there will be practically the majority who use it in their daily lives, probably for reasons of efficiency, in the same way that the vast majority of works and designs today are created through computer programs such as Adobe Photoshop.
First legal-practical approaches to this phenomenon. The case of Zarya of the Dawn
Having made clear the impact of AI on the human creative process, let us now look at an example of one of the first legal-practical approaches to the phenomenon of (AI-assisted works).
The case, undoubtedly mediatic and controversial for some, is that of the comic Zarya of the Dawn and the cancellation of its registration with the US Copyright Office (USCO) on February 23, 20239.
Kristina Kashtanova, a renowned artist and “AI-generated art” evangelist, presented on September 15, 2022 a comic book titled Zarya of the Dawn10, composed (like any other comic book) of images and text vignettes around a storyline. In particular, for the realization of the images, the artist made use of the generative AI tool Midjourney, using hundreds and even thousands of prompts to generate the perfect images. The USCO only acknowledged the artist's authorship (and consequent protection) of the texts, but denied it with respect to the images, as they were not the result of human creation.
The U.S. office, after analyzing the operation of the Midjourney technology and the fit of copyright protection of its resulting images, concluded that users cannot claim authorship of AI-generated images or register intellectual property rights in their name, as they are not the material authors of the images11.
In its reasoning, the Office concludes that, unlike what happens when a user makes use of a tool such as Adobe Photoshop, Midjourney users do not have control over the resulting images, being indifferent -in the Office's opinion- that the user has required hundreds or even thousands of prompts to generate them. It is striking, however, that all the images follow the same style -very characteristic of the artist-, which would not be possible if the user had no real control over what he/she is generating.
What has happened is that practically the entire focus of attention has been centered on the question of non-human authorship, since it is the machine that carries out the expression of the work. In fact, the resolution itself examines the final touches made by the artist using the Adobe Photoshop program (which would be the phase of the creative process that we have previously referred to as 'revision'), and while recognizing the creative effort behind the generation of a particular work, the criterion that has prevailed over any other has been that of non-human expression.
Legal analysis
Before proceeding with the analysis of the issue in the light of continental copyright law, I consider it appropriate to differentiate between the different classes of works generated by AI. Thus, according to the classification made by Professor Pablo Fernández Carballo-Calero12, we can differentiate between (i) works produced by AI systems autonomously and (ii) works produced by AI systems in which a relevant human contribution is verified.
While the first class of works is more typical of a futuristic scenario that does not yet exist today -referring to an Artificial General Intelligence (AGI), we will focus on the second class of works, in order to try to determine whether only on the basis of Prompts we can consider that there can be a relevant human contribution to the creative process, which makes the resulting work worthy of protection by European copyright law.
In this sense, therefore, the question arises as to which of the phases of the creative process produces the germ of originality, which is, after all, what copyright seeks to protect. If it turns out that originality occurs in the phases of 'conception' and 'revision', we will be able to determine, provided the requirements are met, that an AI-assisted work created on the basis of 'Prompts' may be worthy of copyright protection. On the other hand, if we conclude that originality occurs in the 'expression' phase, then the denial of protection will be peaceful, since the status of author should [hypothetically] fall on the AI system, and as we all know, machines, for the time being, do not have the capacity to hold any rights whatsoever.
Following on from the above, and ruling out the possibility of the AI system generating works on its own without human instigation, can Prompts be equated to specific instructions that make a work an intellectual creation of its author13? If we ask ourselves this question thinking about the illustrations of Zarya of the Dawn, it does not seem unreasonable to think of an affirmative answer if we consider the iterative creative process carried out by Kashtanova. In this process, through a succession of Prompts, the author goes back and forth between the phases of 'conception' and 'revision', reflecting in the final illustrations both her free and creative choices14 (see the selection of style, colors and tones used, etc.), as well as her 'personal touch'15 (that which we could qualify, seeing the rest of her work, as Kashtanova's own universe).
The fit of 'Prompts' in the generation of works using AI systems
As practically everyone knows by now, Prompts are our way of telling the machine (in this case, the AI system) what to do. A very simple Prompt, in general, will instigate the machine to carry out 'choices' of all kinds to generate an out-put (which, as we explain in this article, are not properly 'choices'). Instead, it will be through iteration, through successive Prompts with respect to what the machine generates that will result in a given author being able to accurately reflect what he has in mind, according to his style and personality.
To give an example, it will not be the same to simply use a Prompt in the style of the one indicated as nº 1; than to use another one in the style of the one indicated as nº 2, and on the proposed results iterate until achieving the particular expression we have in mind.
Prompt No. 116: ‘a portrait of a robot’
Prompt No. 217: ‘3 d close – up portrait of the retro futuristic robot in vintage death beath mask in mad max style clothes and crown overgrown with fungi and giant orchid flowers and giant gladiola flowers, intricate, elegant, surrounded by smoke and burning. translucent nautilus, highly detailed, by wlop, tooth wu, greg rutkowski, alena aenami’18
It would be something similar to hiring a sculptor to make us a sculpture, differentiating between a simple instruction such as 'make me a sculpture in the shape of a lion', and a succession of complex instructions in terms of shape, style, dimensions, textures, etc. The former would probably lead us to conclude that there has been practically no choice on our part as to the resulting work (but there has been on the part of the sculptor), but in the latter, the response could be more akin to an intervention that could be considered relevant.
The fit of 'Prompts' in the generation of works using AI systems
Following on from the above, I think it is appropriate to bring up a couple of cases that in one way or another support the hypothesis I have been formulating, that the 'expression' phase is secondary if the originality comes from the 'conception' and 'revision' phases.
The first of these is the legal dispute between the famous painter Miquel Barceló and the ceramist/potter Jeroni Murtó, whose conflict was resolved by the Provincial Court of the Balearic Islands in a ruling of January 22, 200819. The facts of interest here are as follows:
The acclaimed painter made use of the technical competence of the potter to carry out a series of ceramic pieces. The material execution of the pieces was carried out by the ceramist, since the painter had no knowledge of pottery. The painter only provided instructions as to style, and modeled and painted the pieces in their final state. The ceramist filed a lawsuit requesting that he be declared as co-author of the works, but the judge considered that it was not proven that «the intellectual creation of the plaintiff is embodied in these ceramics".
Something similar to what we have just seen happened in the dispute between the artist Maurizio Cattelan and the sculptor Daniel Druet, in which the latter claimed in court his recognition as the author of sculptures such as "Him" (Hitler kneeling) or "La Nona Ora", because they had been exhibited by the former, who attributed full authorship to himself. Indeed, who conceived the works and gave the indications for their execution (in a conceptual sense, but not material) was Cattelan, and the judge ruled in his favor as a consequence of the latter by means of a sentence handed down on July 8, 2022 by the Third Chamber of the Court of Paris20.
Conclusions
While until now the differentiation between phases of the creative process was more typical of exceptional cases such as Druet or Baceló, with the relentless advance of generative AI, the need to reach a consensus on what is more relevant to us from a copyright perspective: protecting the result of 'conception' and 'revision', in the European jurisprudential sense of reference21; or just the 'expression'. Perhaps here, as in many other areas of technological progress, we must turn to the philosophical foundations of the field in question22, thus returning to the main questions that once justified the inclusion in laws, treaties and conventions of the attribution of property rights over intangible assets23.
Of course, and going back to this article’s initial question, according to European continental law and the case law of reference of the CJEU, we could defend that on the basis of Prompts there is the possibility of impregnating the work generated through the AI system with our personal touch, verifying in such a case a relevant human intervention that makes such work worthy of protection by copyright. In this sense, as Kalin Hristov points out: "An example may be the creation of a painting by an artist who has selected the colors, tool type (brush size and stroke style) and has to some extent input his requirements into the AI algorithm used to create the work. Although the artist cannot exactly predict the final version of the generated painting, he has directly contributed to its creation and has some expectations as to what it may look like"24.
That is why, in my view, we should be able to look beyond this trend, and leave aside (at least for the moment) the debate about whether or not a machine can have rights, to focus on the rights of the one who instigates the machine, depending on how and to what extent it does so. We will probably have no choice but to resort to utilitarian theories if we do not want to turn our backs on this phenomenon.
In short, it is undoubtedly an exciting discussion, the practical implications of which will most likely be of vital importance for the future of the art, culture and entertainment industry. Whatever the outcome, and regardless of the position we adopt, all this reminds me of the famous anecdote of the French composer Ernest Bourguet (founder of the Société de Auteurs, Compositeurs, et Editeurs de Musique (SACEM), the collective management entity of musical authors in France), who in 1850, while dining in a restaurant, heard how the musicians of the restaurant played one of his works without his authorization or acknowledging his authorship, to which he reacted by getting up and leaving the restaurant without paying as an act of protest.
Footnotes
- Is Netflix a Technology Company? Exploring Its Impact and Role in the Industry, The Enlightened Mindset (lihpao.com).
- See, for example, the following Wikipedia article on 'Artificial intelligence Art'.
- “An AI Prompt is any form of text, question, information, or coding that communicates to AI what response you’re looking for. Adjusting how you phrase your prompt, AI could provide varying responses”, What is an AI Prompt?”, coschedule.com.
- And this is extrapolable to almost any other field of knowledge, see, for example, the evolution of programming languages from binary to high-level programming languages like JavaScript, with a syntax increasingly closer to human cognitive capacity.
- See, for example, "Why tech billionaires like Elon Musk, Bill Gates, and Jeff Bezos are all investing in biotech startups that want to link your computer directly to your brain”, Business Insider India.
- Image No. 1. Diagram extracted from the article Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output? P. Bernt Hugenholtz, João Pedro Quintais. Springer, showing the iterations of the creative process.
- Image No. 2. Labeling of the phases 'conception', 'execution' and 'redaction', based on whether they are phases carried out by humans or by machines.
- Image No. 3. Back cover of issue No. 346 of Món Jurídic magazine.
- Access to the resolution of the US Copyright Office (USCO) through the following link.
- Image No. 4. Cover of the comic Zarya of the Daw (left) and page #2 of the comic (right). ©Kristina Kashtanova
- This is reflected by the Office on numerous occasions: “Courts interpreting the phrase “works of authorship” have uniformly limited it to the creations of human authors”; “the Office “will refuse to register a claim if it determines that a human being did not create the work”. For this purpose, indirect reference is also made to the case of the Naruto Seflie, in as much as it is said: “providing examples of works lacking human authorship such as “a photograph taken by a monkey".
- La Propiedad Intelectual de las obras creadas por Inteligencia Artificial, Thomson Reuters Aranzadi, 2021, pages 64 to 72.
- Case C‑5/08 – Infopaq International v. Danske Dagblades Forening (2009) ECLI:EU:C:2009:465 ((Infopaq)).
- Case C-683/17 – Cofemel — Sociedade de Vestuário, S.A. y G-Star Raw CV (2019) ECLI:EU:C:2019:721 (Cofemel). § 30.
- Case C-145/10 – Eva-Maria Painer v. Standard VerlagsGmbH (2011) ECLI:EU:C:2011:798 ((Painer)), § 98.
- Image No. 5. Example of one of the results generated by Stable Diffusion when using Prompt #1
- Image No. 6. Example of one of the results generated by Stable Diffusion when using Prompt #2
- Prompt No. 2. Retrieved from Stable Difussion prompts search engine.
- SAP nº 24/2008, de 22 de enero de 2008 (ECLI:ES:APIB:2008:106) (Barceló).
- Décisión du 08 Juillet 2022 (Caso Druet y Cattelan).
- We are talking about Infopaq's famous 'author's own intellectual creation', (Infopaq), Cofemel's 'free and creative choices' criterion Cofemel, or Painer's 'personal touch' contribution (Painer).
- The famous historian Yuval Noah Harari, in his work 21 Lessons for the 21st Century, speaks, in his opinion, of the relevance that thinkers, philosophers and historians should take in the search for answers to the great questions of the human being; otherwise, these answers will be given by the engineers who create technology, without their participation of any kind. See, for example, the classic debate on the criterion for prioritizing the risk of the autonomous car.
- This is what Professor Pablo Fernández Carballo-Calero, in his aforementioned work 'La propiedad Intelectual de las obras creadas por Inteligencia Artificial', in the section where he talks about 'Los fundamentos del derecho de autor y su incidencia en las obras creadas por AI' distinguishes between the following three substantive theories: (i) Locke's 'labor theory', which defends property as a reward for the fruit of the work developed by the author, (ii) the 'personalist theory', which places value on the personality of creators, and (iii) the 'utilitarian theory', which examines intellectual property rules in terms of their effectiveness and ability to promote the welfare of society by increasing its cultural heritage.
- HRISTOV, K. “Artificial intelligence and the copyright dilemma".
Leave a Reply