Can Artificial Intelligence Be an Author?

Can Artificial Intelligence Be an Author?

The involvement of Artificial Intelligence (AI) can transcend the outcomes predetermined by a user; consequently, AI itself could be perceived as an author, given that modern AI possesses the capability to create works without human intervention. This theory is quite provocative, as it directly contradicts the standard definition of authorship, according to which an author is a natural person through whose intellectual-creative activity a work is produced. It is important to note that the primary-and perhaps only-advantage of machine authorship is that it aligns with the core logic of intellectual property rights, which dictates that the creator is the author.

Granting authorship to AI may be possible and permissible under specific circumstances, namely: when the work created by AI is not predetermined by a human, when there is no human intervention in the creative process-ensuring the algorithm's independent operation-and when the AI has the ability to independently decide when to create future works. Consequently, it is implied that machines may only be defined as authors when they become autonomous and self-aware at a human-like level. Thus, AI authorship can only be entertained from this perspective; however, from the standpoint of the traditional concept of authorship, this logic remains unacceptable.

According to United States law, copyright protects the original creation of an author, allowing the author to hold copyrights and, through these rights, to adapt, sell, reproduce, and derive material benefit from the work. Furthermore, it should be noted that copyright does not consist solely of economic rights; one of the most significant aspects is authenticity, through which a specific person must be recognized as the author of the work. Accordingly, authorship/inventorship is inherently linked to property rights-rights exercised solely at the free discretion of their owner, a capacity machines do not possess. How would AI decide to whom it should transfer economic rights via copyright? With whom would it execute exclusive or non-exclusive licenses? Or how should the income derived from the work be distributed to it? These are critical questions inherent to copyright that are entirely bound to human agency.

The U.S. Congressional Commission on New Technological Uses of Copyrighted Works (CONTU) noted that AI is a creative tool. It held that “copyright authorship is attributed to the one who designed, programmed, or deployed the machine to generate the resulting work, just as copyright authorship is attributed to the one who used a brush, a saxophone, or a text editor to create a creative work." There was never a question of granting authorship to brushes or other instruments, despite their direct involvement in the act of expressive creation. Therefore, the mere involvement of AI in the creation of a work does not qualify it as an author. A very interesting and, one might say, precedential decision was rendered by a Chinese court regarding a report generated by the Wolters Kluwer software-a legal information tool. The plaintiff claimed that after the AI generated the report, the plaintiff's team further processed and modified it. For the written portion, the team used the tool to find decisions based on the plaintiff’s criteria, then reviewed the search results and excluded irrelevant court decisions. Subsequently, the team analyzed the data and compiled the report. The defendant argued that since the report was entirely generated by AI and thus lacked copyright protection, its use did not constitute an infringement. The court ruled that since there was human contribution and the work was created by a team, it was original and protectable, as the report prepared by the AI reflected choices, judgment, and analysis of relevant data. Beyond this reasoning, the court also addressed the issue of authorship, clarifying that AI cannot be considered an author, nor can the creator of the computer program be considered the author, as the latter did not use the relevant keywords for creating the report and played a minimal role in the work's creation. Additionally, it explained that significant weight should be given to the investment made by the creator of the computer program and to the user who paid a specific fee to use it. Consequently, the court explicitly opposes AI authorship and clarifies that a machine cannot be the creator of a work or a copyright holder, emphasizing that the focus should instead be placed on the roles of the software creator and the user in the creation of such works.

Emphasis must also be placed on the intent to create works in the future, as the concept of recognizing AI as an author contradicts one of the fundamental goals of intellectual property law: providing incentives to authors for further creativity, thereby expanding creative expression and knowledge for the benefit of society. Furthermore, among opponents of AI authorship, creativity holds a special role, a standard under which AI-generated works fail to meet the requirements of creativity and originality. In my assessment, beyond the fact that AI-generated works lack creative potential, AI operates through "templates," creating based on previously supplied or learned information, which further diminishes its creative potential. I would particularly highlight the capacity to own and exercise copyrights, which AI lacks, as a vital issue in copyright is the economic benefit derived from creative activity. It is essential for authors to be independent in their decision-making process and to decide for themselves whether to publish a work through a publisher, lease it, or otherwise generate income from it. Given that AI is not a human and lacks decision-making capacity, it is impossible for it to conduct negotiations or distribute income.

In addition to the exercise of economic and moral rights, the term of protection is crucial for copyright, as these rights operate for a specific duration during which the author may claim royalties. This term varies and largely depends on national approaches, but in most cases, it is calculated as the life of the author plus 70 years-lasting for the duration of the author's life and 70 years after their death. Consequently, this is the specific period during which economic copyrights apply to the work, allowing the author to derive financial benefit. Considering this, there will be several questions: How should the copyright term be calculated if we define AI as the author? How could copyright exist during its existence and for 70 years after its "death"? Is the “death” of AI even possible? These are the significant questions that must be asked and answered when considering the recognition of AI as an author. Furthermore, one of the key issues related to the exercise of copyright concerns the possibility of succession: copyrights are transferred by law, by will (inheritance), or by contract. Unless otherwise provided by a will, the exclusive rights to use the work pass to the heirs. Thus, while copyright allows an author to derive material benefit and use the work at will, it also grants the right to transfer the use of the work to heirs so they may decide how to utilize it. Therefore, even if we assume the aforementioned prerequisites are met-that no human participated in the creation, the result was not predetermined, and the AI acted independently-the issue of the copyright term and its subsequent transfer remains. These are rights entirely dependent on humans, humanity, life, and death. Consequently, beyond creativity and intellectual-creative effort, recognition as an author requires human attributes such as life, inheritance, the right to receive economic income, and authenticity, which are inherently linked to humanity. For this reason, AI cannot be recognized as an author, as such recognition would call into question the very feasibility of exercising copyrights.

In conclusion, while AI is evolving and can produce works independently of a user’s predetermined outcome, this principle remains an anomaly because, under established principles, an author is defined as a human who has invested intellectual-creative effort in the process. In contrast, AI can only be considered an author under specific circumstances: when the work is prepared without a predetermined human result, without human intervention in the creation process, and where there is potential for future works. These are the specific scenarios through which the theory of AI authorship might develop. When discussing AI as an author, it is vital to understand the essence of copyright itself, which enables an author to derive economic benefit through the sale, rental, or transfer of rights. While this is a primary goal of copyright, the issue of authenticity is equally significant, as it recognizes the author as the creator of a specific work. Consequently, AI, being neither human nor a sentient being, cannot realize copyrights. How would it execute a licensing agreement? How would it receive income? These are critical questions that cannot be answered if AI is designated as the author. Moreover, copyright is limited by a specific term-typically 70 years post-mortem-during which material benefits are secured. If AI were the author, how would this term be measured? Life and death are strictly human characteristics. Furthermore, since copyright can be transferred through inheritance, and inheritance depends on the death of a person, this process is entirely human-centric. Therefore, AI cannot be an author because the exercise of copyright and the concept of authorship are fully dependent on human characteristics that AI does not possess. Additionally, the desire and potential for future creation is a fundamental element that cannot be satisfied within the framework of AI authorship.