Who owns AI creation?
Artificial intelligence and intellectual property.Artificial intelligence (AI)-assisted creation poses a real challenge to our intellectual property law system at the national, European, and international levels. In the absence of specific regulations, this new world is governed a priori by the different Terms and Conditions established by the AI companies themselves, which are diverse but have some common features.Property of the input.In general, input refers to the data provided by the user to the AI so it can generate the result, and output refers to the result of the work performed by the AI itself.The terms of use of each platform analyzed establish the same intellectual property regime for the information provided by the user (input). Logically, it is required that the data or information we provide to generate a result cannot infringe third-party intellectual property rights, image rights, personal data protection rights, or violate any other regulations. If the ownership of the data provided belongs to the user, they are obligated to share with the platform unlimited and international rights to use, copy, and modify said data for the achievement of the purposes of said platform. So far, so predictable. If the ownership of the resulting content remains in the hands of the user, they are equally obligated to share with the platform unlimited rights to use, copy, and modify the content for the achievement of the purposes of said platform. Typically, such purposes are related to internal purposes, such as self-promotion and algorithm training.Runway Gen-1, which produces new videos based on user-uploaded content, makes it clear that it has no rights to this original content. Since these are audiovisual inputs, their appropriation would have been difficult to justify. Let's not forget that in their early days, social media, especially Facebook in 2009, attempted to acquire all rights to user-uploaded content through their terms and conditions, but they reversed themselves after the controversy surrounding them.Ownership of output.Regarding the output created with AI, some companies are more generous than others. Open AI, the creator of ChatGPT, grants full intellectual property rights to the user with the only requirement being to clearly indicate co-authorship with the AI in the event of content publication. It is also the only platform that considers the possible appearance of identical results due to the same algorithm. In this case, no one has any rights to the creation.This means that there is no full guarantee regarding the generated result, but I believe this does not imply that third parties, without authorization, could freely use such generated content, because the status of author is not granted to you by an AI platform but by the law. It is somewhat similar to how when two uncopied works coincide in that they are identical or very similar, both would maintain originality or authorship vis-à-vis third parties, but the difficult part is proving that the second work that appeared did not copy the first. Before the internet, it was simpler; it was necessary to demonstrate that, due to context, both authors had difficult access to the other's work. However, currently, there's no need to prove anything; the works match. Consider, for example, a script for an advertisement resulting from a public platform's algorithm that is very similar to that of another agency, and which is also on the red line for plagiarism. Wouldn't the second agency, which coincided with the first when fine-tuning the algorithm, then have any problem? Common sense would suggest that if the first agency aired the advertisement before, the second agency could be considered plagiarism, even if it's the result of the algorithm. In this sense, I think the judges wouldn't be too troubled. And if no agency has aired it before, and both agencies produce it in parallel and independently without any connection, it certainly wouldn't be considered plagiarism. The controversy is there.Runway Gen-1 also grants ownership of the result to the user on the condition that they share the property rights to the content with the company.Taking advantage of the lack of regulation in this regard, Midjourney, which creates new images with a brief description, monetized these different intellectual property regimes. In any case, the company requires users to share the exploitation and distribution rights with the company. If the user uses the platform without an account, they will only be entitled to free distribution of the result. If they have a "Pro" (paid) account, they will retain the intellectual property rights of the result (with the aforementioned requirement). If, in addition to the "Pro" account, they subscribe to the "Stealth" option, Midjourney will agree to self-limit the distribution of the result to avoid overlaps with third parties. This example may give us clues as to where this business model may evolve.More restrictively, the specialized text generator Essay Genius makes it clear that the generated outputs may only be used in the personal context that motivated the use of the platform, for example, in an academic essay.On the other hand, the Tome platform, which allows users to create interesting professional presentations, simply states that the content provided by the platform will always remain its property.Regarding possible infringements of third-party intellectual property, that is, if the results violate such rights, the AI platforms make it very clear that they cannot be held liable under any circumstances. Therefore, they add to their terms that they have no obligation to verify the legality of user-uploaded content, as any criminal use of their services or infringement of third-party rights is prohibited. Regarding generated content, all these platforms use an "as is" clause: what is created by the AI and the user is provided "as is." That is, it is the user who has the obligation to verify that they can use the new content without violating anyone's intellectual property.Summary table, ownership, and exploitation rights. (Own elaboration)In summary, the general trend regarding rights over AI-assisted creations seems to be, on the one hand, to leave moral rights to the user and, on the other, to share exploitation rights. At this point, it would be interesting to see what each platform means by sharing exploitation rights: those of the agency when it sells the Work to the advertiser (ad, designs, texts, etc.), or those of the advertiser when it markets such Works.It is, therefore, the user who will have to verify that they can use the new content without violating the intellectual property of any third party. And this includes intellectual property generated by other AI rights.This "filter" responsibility that falls on the user makes sense when it comes to a creation forged by their own and exclusive will. On the contrary, holding them solely responsible for the production of someone else's will may be excessive.This "filter" responsibility placed on the user makes sense when it comes to a creation forged by their own will. On the contrary, holding them 100% responsible for the production of someone else's will may be ethically and legally questionable. Making AI companies jointly responsible could make such businesses economically unviable due to the high risk of conflict and unpredictability. Of course, in many cases, these AI companies also want to charge for exploitation rights. In other words, the profits are shared, but the risk is for the user.We are at the dawn of this type of creative cooperation between machine and human beings, which will be regulated on a case-by-case basis.And what does the law say?The main reference is the private regulation that each AI company establishes with its users. Nor does current legislation offer many answers when it comes to determining what belongs to whom. As previously explained, no law covers AI-assisted intellectual production. The current framework is limited, but case law has provided us with clues that can serve as a starting point.And the jurisprudence?In Spain, intellectual property is governed by Royal Legislative Decree 1/1996, which establishes "original creations" as the object of protection. That said, not much more is known about what a creation is, and even less about what can be classified as original. Purely digital works are largely left out of the list of types of works. In this area, only databases are explicitly protected by "the selection or arrangement of their contents," which constitute an intellectual creation.The Provincial Court of Navarre attempted to clarify what could be understood by "creative" and "original" to determine whether a production will be protected by intellectual property. It provides two criteria for verifying originality: that it did not previously exist and that it embodies the author's personality. Regarding creativity, the Court equates it with an "intellectual effort" that allows one to move from a simple object to an "artistic or intellectual creation." What's interesting about this ruling is that the work in question was obtained through the use of technological means (a camera), and the existence of copyright does not seem to depend at all on the technique used to obtain the result.The case law of the Court of Justice of the European Union follows, more simply, the same line. In a similar case (another photograph), the CJEU ruled out the relevance of any consideration of the medium used by the person to protect a production, sufficing that it "reflects their personality and expresses their free and creative choices." In short, point the lens and shoot. If the camera takes a photo on its own, for whatever reason, it would not be considered an original work as there was no human intervention. The important thing is that human intervention exists, even if minimal, and this idea can be distorted in the case of AI assistants, because the disproportion between human intervention and the generated result can be abysmal. Thus, if we assume intellectual production generated by AI that does not infringe pre-existing copyright, two possibilities would arise. In the first, the individual's choice of inputs, the selection of outputs, and the manner of presenting them would be considered to reflect the author's effort and subjectivity: AI would be nothing more than a technical means, and humans would be the sole holders of intellectual property. In the second, the final result would be due almost solely to the algorithm within the AI, with the individual being unable to claim anything other than having given it minimal information for it to function. In this case, the production may have an owner who remains to be defined, but it cannot be regulated by intellectual property as we currently understand it, because the algorithm's output has been fed by the prior contributions of thousands or millions of users.Conclusions. More and more could be extrapolated from existing laws and rulings, but these will only be hypotheses. Regulations on this matter are needed, because while the current ones can temporarily make up for the absence of more specific ones, the rapid development of AI will create increasingly more legal gray areas and many controversies. The idea of giving personality to AI does not seem to be, for the moment, the European Parliament's. Furthermore, it would mean recognizing, if not the intelligence of something non-human in the proper and defining sense of the human race, then a creative capacity, which we are not yet ready to assume. In any case, the European Parliament is in favor of a community regulation of the matter that would assign ownership of each work to an author, which is what is understood by author under current regulations. However, the fundamental question remains unresolved: whether 20th-century intellectual property laws are useful for 21st-century AI.Although specific regulation of the matter is desirable, it will not mean the end of conflicts regarding the matter due to the international dimension of the world of AI. On each of the aforementioned companies' websites, the Terms and Conditions clearly state that any dispute related to the use of AI must be resolved by arbitration and, if that fails, by the courts of the state where the company is headquartered, generally California. Therefore, even if there is a desire to regionally regulate AI-assisted intellectual production, its presence and that of the user in different legal systems will lead to further doubts about where and under what law conflicts should be resolved. A real solution would be a new international agreement promoted by the World Intellectual Property Organization to establish basic rules for co-creation with AI without discouraging it. But there is still a long way to go.This article has sought to provide some general guidelines for understanding the legal logic in this type of AI-assisted creation systems. The information provided by the user (input), the result of the system (output), and the liability regime, which generally falls on the user. For this reason, in the next article, we will discuss a series of practical legal tips for using these AI systems in campaign design.Riestra Abogados 2023 Términos y condiciones de las empresas Propiedad de los inputs Propiedad de los outputs Derecho de explotación de los inputs Derecho de explotación de los outputs Open AI (ChatGPT) Usuario Usuario Si respuestas idénticas: nadie Usuario Usuario (con mención “hecho con ChatGPT”) Tome Usuario (impreciso) Usuario y empresa Usuario y empresa Midjourney Usuario Gratis: empresa Usuario Gratis: usuario (sólo distribución) y empresa Pro: usuario Pro: usuario y empresa (limitable) Runway Usuario Usuario Usuario Usuario y empresa EssayGenius Usuario Usuario Usuario Usuario (únicamente para un uso particular