Can artificial intelligence be the author of a film work?
Artificial intelligence has experienced significant developments in the past decade, thanks to the massive amount of available data, improved computer performance, and the development of new architecture based on deep learning. In contrast to a classic computer program which represents a simple tool for work in the hands of a person and older artificial intelligence systems based on behaviour prediction, new forms of artificial intelligence which are based on the principles of deep learning or artificial neural networks are capable of autonomously generating completely new content (hence the name generative artificial intelligence).
We all feel the consequences of the so-called ’fourth revolution’, whether we are excited about the new possibilities of applying artificial intelligence or anxious about the potential job loss. Artificial intelligence arrives in many forms and with different purposes, along with the potential to replace human work in any industry. The creative industry has proven itself to be particularly fertile ground for the "proliferation" of various forms of artificial intelligence. Artificial intelligence composes music, writes songs, novels, newspaper articles, creates pictures and photographs, and recently began to "film" movies!
Currently, the most famous artificial intelligence model for creating audiovisual works is SORA. Developed by the company Open AI, this model generates short video clips based on user prompts (text-to-video model). Several other models for generating audiovisual content from prompts were created before SORA, including Make-A-Video by Meta and Google's Lumiere. That artificial intelligence can also be a good screenwriter was demonstrated by the 2016 Sci-Fi film "Sunspring", which made it to the top 10 at the annual Sci-Fi Film Festival in London. The complete script was written by Benjamin AI which is based on an artificial neural network and trained on the scripts of sci-fi movies from the 1980’s and 90’s.
Given that generative artificial intelligence creates completely new content which is difficult to relate to the intellectual work of a specific natural person, whether it is the developer of the artificial intelligence system, the author of the work that was used for its training, or the user who initiated the creation process with his instructions, the question of the position of such content within the current system of copyright protection has arisen, or specifically for us in this text the question of whether artificial intelligence can be considered the author of a film.
The answer to this question can be found in the Law on Copyright and Related Rights of Republic of Serbia, specifically in Article 2, which defines the concept of copyrighted work, and Article 9, which prescribes who is the author of the work. In other words, we must first answer the question of whether the content created by artificial intelligence is subject to copyright, and only if the answer is affirmative, who is the author of that work.
Article 2, paragraph 1 of the Law on Copyright and Related Rights defines copyrighted work as the original intellectual creation of the author, expressed in a certain form, regardless of its artistic, scientific or other value, its purpose, size, content and manner of expression, as well as the permissibility of publicly communicating its content.
Therefore, in order for a work to be copyrighted, it must represent the original spiritual expression of the author, an expression of their personality, creation, state of mind, i.e. intellect. Originality and spirituality are two sides of the same coin since originality is defined by the representation of the author's personality. And already at this first step, artificial intelligence has failed the test, given that the content it generates cannot represent an expression of its personality (because this is a quality it does not possess), and it does not represent their spiritual creation, at least until the moment when artificial intelligence becomes "real " intelligence, or until it becomes a self-aware being capable of representing "itself" through its work.
The legislator is also aware of this, therefore Article 9 of the Law on Copyright and Related Rights stipulates that an author is a natural person who created the copyrighted work. Therefore, only a natural person can express the state of their spirit in a certain form, i.e., only they can create a copyrighted work. Artificial intelligence is not a natural person, and in accordance with current regulations, it will not be considered the author of the content it has generated. Considering that artificial intelligence is not even a legal entity and that the question of responsibility for the damage caused by it acting autonomously in accordance with its own programming and purpose could potentially be raised, the European Parliament is considering the possibility of introducing a third category of persons, so-called "electronic persons", which would include robots and artificial intelligence.
To conclude, in accordance with domestic positive legal regulations, content generated by artificial intelligence does not represent work that can be attributed to the authorship of artificial intelligence by itself.
And, even if it would seem logical at first glance, we did not quote Article 11 of the Law on Copyright and Related Rights which states that the authors of a film work are the screenwriter, the director and the chief cinematographer for pushing our case, because the question whether artificial intelligence can be a director or screenwriter is actually already contained in the primary question of whether artificial intelligence can be an author. If it can be an author, then it can also be a screenwriter, so Benjamin would be considered the author of the award-winning film work "Sunspring", which is certainly not the case.
The inability of artificial intelligence to be considered the author of the content it generates leads to two possibilities: either the work produced by artificial intelligence falls into the public domain, or another individual is recognized as the author of that work.
Artworks in the public domain are works on which no person has the right of ownership, i.e. they are part of the general cultural and intellectual heritage of humanity that anyone can use. The question arises whether, from the point of view of "incentive theory", it is at all necessary to extend the system of copyright protection to works generated by artificial intelligence, or whether it is more rational to consider them as work in the public domain. Also, even if we assume that these artworks are not in the public domain, it may be possible to protect them with other legal mechanisms such as the ’right of the first publisher’ (publication rights) or the creation of a new related right similar to database producer rights (which would de facto protect the investment in the artificial intelligence system). Certainly, all of these issues should be explored from both a legal and an economic point of view before favoring any solution.
On the other hand, the authorship of a work generated by artificial intelligence could eventually be attributed to a natural person. Who that person could be depends on the extent of creative influence on the final product, and potential candidates are primarily the programmer who created the artificial intelligence system and the user based on whose instructions (prompts) the content creation process was initiated.
The programmer of the artificial intelligence system could appear as a possible author, given that artificial intelligence is based on a computer program, the author of which is a natural person - the programmer. However, the problem with this solution is that, at least when it comes to new forms of artificial intelligence (deep learning and artificial neural networks), there does not seem to be a direct causal link between the software developer and the final generated result, i.e. those forms of artificial intelligence which are able to generate completely new content. There is already judicial practice on this issue, therefore in the case of Rearden vs. Walt Disney Co., the American court took the position that films made using Rearden's software cannot be considered derivative works of the software because even though the software did a significant part of the work, the decisive share of creative expression in the film can be attributed only to the employees of the Walt Disney Company.
Content generated by artificial intelligence is rather an expression of instructions or prompts given for the creation of certain specific content, so recognizing the authorship of the user of artificial intelligence could be a more acceptable solution. However, we must bear in mind that the mere selection of data and its classification by the user is not enough to fulfil the requirement of "originality", rather recognition of authorship to the user would be possible only in the case of his real and essential contribution to the management of the artificial intelligence system in the process of creation, that is, if the key creative choices are made by the user and not by artificial intelligence. Therefore, to the extent that the content generated by artificial intelligence is an expression of the user's creative spirit or the message that they wanted to convey to their audience and to the extent that the user can foresee such a result, such content would be his work as an author. In the event that the generated content is the creative result of both human and machine choices, content that is the result of choices made by artificial intelligence must be exempt from copyright protection. Only independent human contributions are protected by copyright.
Finally, the influence of the work of authors which has been used for artificial intelligence training also cannot be ignored, however, as we are dealing with millions of works by known and unknown authors, it would not be feasible to include them as potential candidates for authorship of the generated content.
We hope to have provided fresh insights into the potential issues and challenges that intellectual property law may encounter in relation to artificial intelligence. Our intention here was to point out the complexity of the question of authorship in relation to works generated by artificial intelligence, not to offer a definitive solution. Finally, it would be interesting to see what the possibilities of generative artificial intelligence really are, so we asked ChatGPT to write the conclusion for us. When we asked what the conclusion of this text would look like, we received the following answer:
"The conclusion of this paper can be summarized in the following points: Authorship of works generated by artificial intelligence presents a challenge to the existing legal framework, because artificial intelligence cannot be an author by itself. Potential authors can be the developers who developed the technology and the users who directed its operation. A key aspect of crediting the user is his significant creative contribution during the content generation process. Also, the rights of the authors of works used to train artificial intelligence should be taken into account, which further complicates the situation. Further research is recommended to determine the best legal solutions that would enable the protection of rights and encourage innovation in this dynamic field."
Author: Stevan Pajović, Lawyer at TS Legal
GRATA International, Serbia
