Sofía Ruiz de la Viuda, Intellectual Property, AI, and Software Consultant at PONS IP

We all remember the 2D “drawing” depictions that emerged before the summer, where by providing a photo to the ChatGPT image generator, it turned us into a version of ourselves that could have come straight out of a traditional Studio Ghibli film set in the Hokkaido Prefecture. But of course, we were at home and no one had drawn us. Our animated version was generated by a probabilistic model based on thousands and thousands of images that had already been processed, ingested and reproduced.

Now, although they have been a long time coming and although the fashion for these Japanese-style 2D “drawings” seems to have passed, the Japanese organisation CODA, which brings together a large number of companies involved in the creation of video games, anime, manga and other entertainment content, including Studio Ghibli, has submitted a letter to OpenAI demanding that its new video generation model, Sora 2, NOT use their works to train this system. This letter could certainly be considered a cease-and-desist order, although it is a very preliminary step before these cases go to court. The truth is that CODA is not alone in its complaints and demands regarding how these AI system providers have handled third-party rights. It is also true that they have not gone as far as other authors’ representatives, given that these large technology providers of generative AI models are also facing lawsuits in the US, Europe and Asia, led by large companies, managers and copyright holders, including the cases of

The Walt Disney Company & Universal Pictures v. Midjourney, Inc., SNE – Syndicat National de l’Édition / SNAC – Syndicat National des Auteurs et Compositeurs / SGDL – Société des Gens de Lettres v. Meta Platforms, Inc., and Sony Music Entertainment and Warner Records Inc. v. Suno Inc. 

While these cases are being resolved, the agreement reached in the Anthropic case (in which a historic out-of-court settlement has been reached with a group of authors and publishers: $1.5 billion, $3,000 for each of the 500,000 works included in the class action lawsuit) may set an interesting precedent, given that this solution involves sharing the profits obtained by technology companies with the authors. The use of generative AI is now commonplace, but that should not make us forget that authors should be offered fair compensation when these models have been fed with their works. Therefore, the questions remain: How do I comply with intellectual property legislation and more general regulations when using these tools? How can I ensure that the use of my inputs and outputs is legal?

On the one hand, and to comply with intellectual property legislation, it is essential to be able to guarantee that the materials we use as a basis for requesting something from an AI tool do not infringe third-party rights. Therefore, we must not upload, copy or use other people’s works, such as texts, illustrations, photographs, music or code, without sufficient authorisation or licence, unless the works are in the public domain or we can be sure that they fall within one of the exceptions provided for by the legislation of each territory (such as data mining in Europe or fair use in the United States).

As for the results generated, current Spanish and European legislation is based on a clear principle: creations produced by AI without significant human intervention are not considered works susceptible to copyright protection. At the same time, and even taking into consideration that the criteria for determining when we are dealing with machine-generated content and when we are dealing with a human creator aided by technology are still being defined, it may be that if the human contribution to the creation provided by an artificial intelligence model is considered substantial and relevant, this could result in copyright being granted on the final result.

By the same token, the problems do not end there, because even if authorship can be proven, the result may still be lacking elements protected by third-party rights. AI models learn from large volumes of data, and sometimes their responses can reproduce, in a more or less obvious manner, fragments or styles of pre-existing works, as is clear from all the controversy generated around the characters created in the style of Studio Ghibli’s drawing. Therefore, if a text, image or generated piece substantially resembles a known work or a copyrighted character, their public use could constitute an infringement. In this regard, it is advisable to carry out manual reviews, keep step-by-step evidence of the creation process (prompts, iterations, modifications) and, in case of doubt, seek legal advice or additional licences from the authors of the original work.

Finally, when it comes to the uses to be made of the results rendered by an AI tool, the restrictions do not end there.  Every provider, whether it’s OpenAI, Midjourney, Runway, etc., establishes specific terms of use that determine the scope of rights over outputs. Some will grant ownership to the user, while others impose restrictive licences or reserve certain rights of use. It is therefore essential to read the terms of service and verify that the intended use of the result is allowed, particularly if that use is expected to be commercial.

From a broader regulatory viewpoint, it should also be taken into consideration that the legality and compliance with legislation when using AI tools is not limited to respecting copyrights. Emerging artificial intelligence legislation, particularly the European Regulation on Artificial Intelligence, which is already being progressively implemented, imposes obligations of transparency, accountability and traceability. Among others, it requires us to report when content has already been generated or altered by automated means, and to maintain human oversight mechanisms not only in the development, but also during the use of AI tools.

Ultimately, the legal use of generative AI tools is based on three pillars:  using only materials for which you have rights or authorisation, known and complying with the terms of use of each platform, and acting with transparency and diligence with regard to the results generated. The key is to document the creative process, verify associated rights, and assume that, in the eyes of the law, ultimate responsibility lies with the person who decides how to use and disseminate the content produced.

Article published in VOZ POPULI, November 17, 2025. You can access the original content here