Just yesterday, I was speaking with an acquaintance from the IT industry. Let us call him Petro (the name has been changed, although the situation itself is entirely real). He told me that his company was effectively sending him on a kind of “creative leave” with a rather straightforward message: it is time to improve his AI-related skills and consider how these tools can generate additional value for the company.
At the same time, there is another story. A friend of mine — let us call her Olha — works at a large IT company and is already actively using AI in projects. According to her, the amount of work has not decreased, but its nature has changed. There is now considerably more time devoted to management, task-setting, and decision-making. Part of the technical workload has effectively been delegated to tools that, only a few years ago, appeared experimental. These two stories accurately reflect what is currently taking place.
Artificial intelligence no longer merely assists — it participates directly in the creation of what has traditionally been regarded as the result of intellectual and creative activity: texts, software code, analytical materials, and more. At a certain point, however, the issue of efficiency is accompanied by another question — authorship. If a result is created not directly by a human being, but with the assistance of AI, a straightforward legal question arises: who owns it?
Ukrainian law is based on the fundamental principle that only a natural person may qualify as an author. Accordingly, AI itself cannot be recognized as an author. However, this explanation is no longer sufficient.
In practice, the final result does not emerge automatically. There is a person who formulates prompts, works with the generated response, analyses and edits it through the prism of professional expertise, and subsequently uses it. This gives rise to the key legal question: does such involvement constitute a sufficient creative contribution?
If a text is merely generated and used “as is,” it is difficult to speak of copyright protection under the current legislative framework. However, where a person substantially shapes the content and final form of the output, elements potentially eligible for legal protection begin to emerge. The difficulty lies in the fact that there is currently no clearly defined legal boundary, and it will most likely develop gradually through judicial practice and regulatory interpretation. For now, it is worth examining both the current legal landscape and the direction in which it may evolve.
A seemingly logical argument is frequently raised: if a person reads books, analyses information, and on that basis creates their own works, how is this fundamentally different from the operation of AI? From this perspective, the distinction may appear minimal, and this forms one of the principal arguments advanced by proponents of extending copyright approaches to AI-generated content.
The underlying logic is relatively simple: if the resulting output is new, does not constitute direct copying, and is created on the basis of pre-existing data, why should it not qualify as a creative work, regardless of who — or what — generated it? At first glance, this position appears persuasive. In practice, however, the distinction remains significant.
A human being works with a limited volume of information and produces a result through personal interpretation, experience, and context. AI operates differently: it processes vast datasets through statistical modelling and probabilistic relationships. The key issue is not intelligence itself, but rather the scale and method of information processing.
Where millions of texts are used for training and subsequent generation, it becomes difficult to compare such activity to individual human learning. Instead, it resembles systematic use of third-party content. For this reason, the legal debate is gradually shifting away from the question of “whether this constitutes creativity” toward another issue entirely: whether such a model of content generation is lawful from the standpoint of copyright law. This is precisely where the real disputes begin.
This is no longer merely a theoretical discussion. The lawsuits against OpenAI, Meta, and Google are fundamentally about the same issue: whether third-party content may lawfully be used to train AI models without authorization. At the same time, no unified judicial approach or settled practice has yet emerged.
In 2025, a U.S. federal court ruled in favour of Meta in litigation initiated by a group of authors concerning the use of their books for training the LLaMA model. The court concluded that, within the specific circumstances of that case, such use fell within the scope of the fair use doctrine, particularly due to the absence of proven adverse impact on the market for the authors’ works. However, this should not be interpreted as a universal authorization or a blanket approval.
In essence, the court merely established that infringement had not been sufficiently proven in that specific dispute. Under different factual circumstances, the outcome could be entirely different.
At present, courts are not prohibiting AI as a phenomenon. Rather, they are attempting to delineate the boundary between lawful training and copyright infringement — and that boundary remains fluid.
Another illustrative example is Thaler v. Perlmutter, where it was expressly confirmed that a work created without human involvement is not eligible for copyright protection. Yet this represents only part of the broader picture. The principal legal conflict today concerns not so much who qualifies as the author, but rather how the resulting output was generated in the first place.
At present, there is no unified global “law on AI and authorship.” However, this does not mean that no regulatory framework exists. In Europe, the EU AI Act is already in force, representing the first comprehensive attempt to regulate this sphere systematically.
Importantly, the regulation does not directly answer the question of authorship. Instead, it establishes principles that may prove even more significant in practice.
The regulatory approach is risk-based: the more complex and sensitive the use of AI, the more stringent the compliance and oversight requirements become. Transparency constitutes another key principle — users should understand when content has been created or modified using AI. A third principle concerns accountability, which extends not only to developers, but also to businesses deploying these technologies. Finally, there is the principle of explainability and control.
The discussion is no longer about a “black box,” but rather about processes that must be understandable, traceable, and reproducible. One of the most interesting yet unresolved practical questions is how exactly to document the creation process in order to demonstrate sufficient human involvement such that the resulting output neither infringes nor unlawfully restricts the rights of third parties, while simultaneously qualifying as a copyright-protected work.
Ukraine does not yet have established judicial practice in this area, but the legal framework is already evolving in this direction. In particular, Article 33 of the Law of Ukraine “On Copyright and Related Rights” introduces a sui generis right applicable to non-original objects generated by a computer program.
In essence, this represents a compromise approach: such outputs are not recognized as classical copyright works, yet they are granted a separate legal regime. This constitutes the first indication that certain categories of AI-generated content no longer fit within the traditional concept of authorship.
At the same time, by Resolution of the Cabinet of Ministers of Ukraine No. 1556-r dated 2 December 2020, the Concept for the Development of Artificial Intelligence in Ukraine was approved, establishing the basic principles and strategic directions for the development of this field.
Although these documents establish a regulatory framework, they do not provide answers to the practical legal issues businesses are already facing. This means that, as in other jurisdictions, legal practice will inevitably evolve faster than regulation itself.
It is also worth noting that Ukraine is already seeing the first examples of legal protection for collaborative human-machine creative results. The Ukrainian Intellectual Property Office has registered copyright in works containing AI-generated images, albeit with an important qualification: legal protection extends only to the portion created by the human author.
If simplified, the legal model currently emerging is already relatively clear. AI may — and should — be used. However, copyright does not arise in favour of AI, nor does it arise automatically. It belongs to the person who meaningfully shapes the final result.
The same applies to registration: almost anything may formally be registered, but in the event of a dispute, the decisive factor will not be the mere existence of an entry in the register, but rather the substantive nature of the work and the extent of human creative contribution.
As for disclosure obligations regarding the use of AI, Ukrainian law currently imposes no such requirement. Nevertheless, the overall regulatory direction is evident: transparency obligations will continue to increase.
The legal risks already exist, yet the absence of unified approaches and clearly defined legal consequences still creates the illusion that such risks may be ignored. In my view, however, companies intending to operate on global markets while actively integrating AI into their products should already begin developing AI compliance strategies — not merely to avoid potential sanctions, but to position themselves among those already aligned with emerging global standards.