On the trail of the creator – the trouble with copyright protection of the effects of AI activity

Legal protection of works created by artificial intelligence (AI) is becoming increasingly controversial and complex. As AI technology develops and finds application in various fields, questions arise about who owns the copyright to the content generated by the algorithms.

Can AI be considered a creator? 

In the light of the applicable law, it should be stated that the creator, in the strict sense of the word, i.e. the entity for which, in principle, copyrights are created, can only be a human being[1].

The above is confirmed by the jurisprudence – the Supreme Court decided that the original source of copyright is the creative activity of a natural person. It follows from the very nature of the creative process that the creator can only be a human being[2], and the features determining the protection of the work can only be given by one. Only the human being in question is equipped with the ability to create or generate a work[3].

The current legal status in Poland does not allow for the recognition of AI as the creator of works, and moreover, among the subjects of civil law, only a human being can perform a creative act[4]. However, this topic is dynamic and requires further research and regulation to adapt the law to the rapidly developing technology.

To whom the authorship of the works generated by the cooperation of AI and man should be attributed?

One of the problems with products created in whole or in part with the participation of AI in copyright law is the answer to the question whether they are protected at all. On the basis of the solutions existing today, it is not possible to attribute them directly to AI, but the authorship can be granted to a person who designs or employs an artificial intelligence system[5] and takes an active part in the process of creating a work using AI.

For example, Image-to-image translation is a method in which a person provides an image that the AI transforms into another one, while maintaining the essential features and structure of the original. However, the use of this tool in the creative process will not be sufficient to assign authorship to the supplier. Here, we can compare the situation where a client employs an artist and provides them with comments on the desired work, what it is supposed to represent, its style, mood, etc. The fact that the client gives the creator tips based on which the client then recognizes the final effect, the image, as corresponding to a certain imaginary concept, does not make them the author of the work[6].

The situation in the USA is also worth a closer look. The local Copyright Office (USCO) has indicated what is decisive for assessing whether the work was created by a human.
Namely, it is important to determine whether the AI served only as an auxiliary tool, or whether the traditional elements of authorship in the work – literary, artistic or musical expression or elements of selection, arrangement, etc. – were in fact invented and made not by man, but by machine.

If the work was created by a machine or a process that works randomly or automatically – without any creative contribution or intervention of a human author – then we cannot talk about human authorship. However, if, during the creation of a work in the program, a person makes certain decisions regarding the features of the work and the creative process, they actively shape their choices in real time during the operation of the program (segregation, selection of elements), and the effect of these choices is created only with the help of the program, such a contribution can be considered creative.

How to distinguish whether the work was created by AI or a human? – overview of methods

The distinction between works created by AI and those created based on human creativity is becoming more and more complicated. There are several key aspects that can help identify the source of the work:

• labelling – aims to ensure transparency and authenticity of materials published in various media. The labels inform the recipients that the material was created or modified using AI tools. An example is the Meta platform, which labels the content generated by AI to promote transparency and understanding among users[7];
• detection using AI – refers to the processes of identification and analysis of content that was generated by AI systems using programs created for this purpose;
• hashing – a technique that involves transforming input data (e.g. text, image) into a unique string of fixed-length characters, called a hash. Each file or piece of content generated by AI can be assigned a unique hash, which allows its identification and verification;
• fingerprinting – it can take into account various data features, such as structure and content, and is used for pattern recognition;
• metadata – they provide information about the origin, author and technical details of the created materials.

The problem of identifying the origin of AI content in light of the provisions of Regulation (EU) 2024/1689 of the European Parliament and of the Council – Artificial Intelligence Act (AIA)

Article 50 (2) of the AIA imposes an obligation to label artificial content in a machine-readable format. This order applies to AI system providers, i.e. systems generating synthetic content, sounds, images, video, text.

An exception to the obligation to label is the operation of auxiliary AI systems, e.g. in the scope of standard editing.

Entities using an AI system that generates synthetic content that can give the impression of being real (deepfakes) must, in principle, disclose that the disseminated content was artificially generated, or possibly subjected to other types of manipulation[8].

[1] M. Markiewicz [in:] Komentarz do ustawy o prawie autorskim i prawach pokrewnych [w:] Ustawy autorskie. Komentarze. Volume I, ed. R. Markiewicz, Warsaw 2021, art. 8.

[2] Judgment of the Supreme Court of 14 Feb 2014, II CSK 281/13, LEX No. 1459159.

[3] Judgment of the Supreme Court of  6 Oct.2017, II FSK 2462/15, LEX No. 2393123.

[4] A. Niewęgłowski [in:] Prawo autorskie. Komentarz, Warsaw 2021, art. 8.

[5] https://zaiks.org.pl/artykuly/2023/grudzien/czego-tworcy-oczekuja-od-ai-act

[6] https://www.traple.pl/tworczosc-ai-a-prawo-autorskie-w-usa-rozstrzygniecia-us-copyright-office-i-wnioski-dla-polskiego-prawnika/

[7] https://www.meta.com/pl-pl/help/artificial-intelligence/how-ai-generated-content-is-identified-and-labeled-on-meta/

[8] D. Flisak, Akt w sprawie sztucznej inteligencji, LEX/el. 2024.