Artificial intelligence is transforming content creation, and is already being applied to both texts and artistic works. But who is the real author when a machine generates something? The debate is not new, as the old paradox of Theseus’ ship already questioned the authenticity of an object if all its parts are changed. Today, it is AI that leads us to ask how much a work must be transformed before it ceases to be a ‘copy’ and becomes something new.

The answer is complex, and also depends on the culture in which we find ourselves. For example, in countries such as China, where imitation has long been valued as a process of learning and homage to the author, innovative elements that did not belong to the original work are often added, a phenomenon with its own name, known as shānzhài (山寨). In Japan, too, there is the case of Nintendo’s success, which often relies on the reuse of technology already developed for other fields. While copyright answers may be more flexible, they are not free of dispute and courts also deny copyright to works generated entirely by algorithms (Wang, 2023).

The Grey Areas: Plagiarism or Reinvention? 

With or without AI, between plagiarism, homage, inspiration and reinvention, it is easy to find grey areas and blurred boundaries. Generally speaking, the first is copying without transformation and without attribution of authorship, while the other three are based on the recognition of having taken the author’s ideas as a basis and having added something of one’s own, something original. In order to also be considered new, the transformation must have been significant.

The recent rise of AIs that generate Studio Ghibli-style images is an excellent example to reflect on: is it plagiarism or algorithmic reinterpretation? The difference with Nintendo’s recontextualisation lies in the AI’s lack of creative intent, which is limited to statistical reprocessing of the works produced by Miyazaki’s studio, in many, if not all cases, without Miyazaki’s prior consent to use the data as training material.

This brings us to the legal field. On the one hand, the challenge is to determine when there is sufficient transformation. As Tanaka et al. (2024) point out,  the question is not only “who created it?”, but also “what added value is there?”, as well as the use made of this content, which, although freely available on the internet, should not be used for profit without due credit and appreciation. While in some cases such content is subject to different types of licensing with regard to its transformation and commercialisation, in many other cases it is still a legal blind spot. 

The line between stolen and reinvented remains blurred; as Martínez (2023) points out, ‘an algorithm can generate a portrait in Picasso’s style, but it does not decide to pay homage to the artist’. The judgment in the case analysed by Cheng (2022) denied copyright to an AI-created work because it required human authorship, similar to the case analysed by Wang (2023), which emphasised the need for human intervention for legal protection. 

Solutions for a Hybrid Age

At this point in time, it is important to reflect on whether it is really possible to speak of reinvention without cultural understanding and artistic intent, as well as on the ethics of using certain data to train a model without the authors’ express permission.

If it is assumed that, like Theseus’ ship with new timbers, AI rebuilds rather than copies, adapted legal models become necessary. Some suggestions include:

• Layered licensing, distinguishing human input from automated input.
• Application patents, which protect innovative uses of existing technologies, as Nintendo has done.
• Contribution labelling to clarify whether a piece of content is human, collaborative or 100% algorithmic.
• Source registration, to keep track of how the original data is transformed.
• An accelerated public domain system, as proposed by Yamamoto (2023), offering short-term protection for AI works that are not authored in the traditional sense.

Conclusion  

Theseus’ ship is now sailing in digital waters and while there are examples of how reinvention can be as valuable as creation from scratch, artificial intelligence clearly challenges notions of authorship and originality and leads us to reflect on and adapt ethical and legal frameworks. The question, really, is how to assign value (and how much) to what is born of a different process.

References  

Chen, Y. (2022). The concept of originality in the copyright issue of AI-generated works in China. Queen Mary University of London.

Martínez, L. (2023). Difuminando fronteras: cómo la IA redefine la propiedad artística. ResearchGate.  

Tanaka, K., et al. (2024). La percepción humana de las creaciones de IA. Springer.  

Wang, H. (2023). Derechos de autor en obras generadas por IA: el fallo judicial chino. Law.asia.  Yamamoto, T. (2023). New AI and copyright challenges. Allatanys.jp.