Artificial intelligence as a subject or tool of creative activity: problems of determining authorship and ownership of exclusive rights
Generative artificial intelligence can produce texts, images and other outputs with varying levels of human involvement. This raises two connected copyright questions: whether a human author can be identified and who holds the exclusive right. The article examines these issues under Part IV of the Civil Code of the Russian Federation and compares selected approaches in the United Kingdom, the United States and China. It distinguishes authorship from ownership of the exclusive right, including in employment and contractual settings, and considers the 2026 Russian AI law. The author argues that AI is a tool, not an author, under current Russian law. Copyright protection depends on identifiable human creative contribution; fully autonomous outputs may require a separate, limited economic right that does not confer authorship.
artificial intelligence, copyright, authorship, author, exclusive right, generative AI, intellectual property, human creative contribution, right holder
- Introduction
Generative artificial intelligence is now used to create texts, illustrations, music, video and other content. The result may be produced after a single prompt, or it may appear after a long process involving the selection of a model, the development of instructions, repeated adjustments, selection among several versions and further editing by a person. These situations are technically similar, but legally they may be very different.
The central problem is not only whether an artificial intelligence system can be called an author. It is also necessary to determine whether the result is protected by copyright at all and, if it is protected, who holds the exclusive right. The user, developer, service provider, employer and customer may all participate in the process, but their participation does not have the same legal meaning.
Russian civil legislation proceeds from the idea that authorship belongs to a citizen whose creative work has produced the result. At the same time, the exclusive right may in certain cases belong to another person, including an employer or a person to whom the right has been transferred. Consequently, the terms “author” and “right holder” should not be used as synonyms.
The purpose of this article is to examine the legal status of AI-generated and AI-assisted results, identify the criteria for human authorship, and consider how the exclusive right should be allocated. The main focus is Russian law, while foreign legislation and judicial practice are used for comparison.
The objectives of the article are:
1) to examine the legal meaning of authorship and the role of human creative contribution;
2) to determine whether an AI system may be recognised as an author under Russian law;
3) to distinguish authorship from ownership of the exclusive right;
4) to compare selected approaches in the United Kingdom, the United States and China;
5) to assess the significance of recent Russian legislation and relevant Russian judicial guidance;
6) to formulate proposals for the treatment of fully autonomous AI-generated outputs.
- Material and methods
The research draws on Part IV of the Civil Code of the Russian Federation and Federal Law No. 243-FZ of 26 July 2026 [1]. It also examines academic works on copyright and artificial intelligence, the UK Copyright, Designs and Patents Act 1988, and the UK Government’s 2026 report on copyright and AI. In addition, the study takes into account materials published by the United States Copyright Office, Russian judicial guidance and relevant court decisions from the United States and China.
The formal legal method is used to interpret the provisions concerning authorship, exclusive rights and works created in the course of employment. The comparative legal method makes it possible to identify differences between the selected national approaches. The analysis of judicial decisions is used to examine how courts assess evidence of human creative participation. The system method is applied to distinguish the legal status of the author, the exclusive right holder and persons who provide technical, organisational or financial assistance. The study is doctrinal and comparative in nature; it does not claim to provide an empirical assessment of the frequency of AI-related copyright disputes.
Artificial intelligence as a subject or a tool of creativity
Article 1228 of the Civil Code of the Russian Federation provides that the author of a result of intellectual activity is a citizen whose creative work has produced that result. It also excludes persons who have provided only technical, consulting, organisational or material assistance, as well as persons who have merely supervised the work. Article 1257 applies the same human-centred approach to works of science, literature and art [2].
Under these provisions, an AI system cannot be recognised as an author. The law refers to a citizen, while an AI system is neither a citizen nor an independent participant in civil legal relations. Its ability to generate a text or image that appears creative does not change its legal status. It also cannot exercise the personal non-property rights connected with authorship in the way provided for by the Civil Code.
This does not mean that every result created with the help of AI must remain outside copyright protection. A camera, editing programme or other technical tool does not become an author simply because it takes part in the creation process. The main question is whether a person has made creative choices that can be identified in the expressive form of the final result. If this condition is met, AI may be treated as a tool used in the creative process rather than as a substitute for the author.
Kalyatin examines possible approaches to determining the person entitled to rights in results created with the use of AI, including the user, investor and organiser of the process [3]. Chernyshov distinguishes between results created with human creative participation and outputs generated without such participation. He argues that the former may be protected under the existing human-centred model, while the latter require a legislative solution [4].
Utkin takes a more restrictive position and concludes that a result created autonomously by artificial intelligence does not satisfy the current criteria of Article 1228 of the Civil Code and therefore is not protected as an object of copyright under the existing model [5]. This position reinforces the need to distinguish AI-assisted works, in which human creative choices may be identified, from fully autonomous outputs.
The positions discussed above point to a more specific question than whether AI should simply be described as a tool. The decisive issue is the extent to which a person’s creative contribution can be traced in the final expression and whether that contribution is sufficient to identify an author.
Human creative contribution as the main criterion of authorship
The most difficult issue is determining what level of human involvement is sufficient for a person to be recognised as an author. The fact that a person has entered a prompt does not automatically prove authorship. A prompt may contain a detailed description of the desired result, but it may also be a general instruction that leaves almost all expressive decisions to the system. The number of prompts, the time spent using the model or the fact that the user paid for access should not be treated as decisive criteria.
In my view, the decisive factor is how far a person’s choices shape the expressive form of the finished work. The assessment may take account of three circumstances. First, did the person develop a specific creative concept and choose the elements through which it was expressed? Second, did they exercise meaningful control over the composition or structure by selecting, arranging or editing the material? Third, can those choices be demonstrated through the history of creation, including drafts, successive versions, prompt records or later edits? These factors are intended as practical guidance; they do not constitute a closed legal test.
A similar distinction appears in the 2025 report of the United States Copyright Office. The report recognises that protection may be available where a human determines sufficiently important expressive elements, for example by contributing original material or making creative modifications and arrangements. The use of prompts, taken on its own, will not necessarily establish authorship [6].
The Beijing Internet Court’s decision in case (2023) Jing 0491 Min Chu No. 11279 provides a useful illustration. In considering the image at issue, the court looked at the user’s selection and arrangement of prompt words, the parameters chosen, and the changes made before the final version was selected. On those facts, it found that the user’s intellectual input and individual choices were reflected in the image, and recognised the user as its author [7]. The decision should not be read as granting copyright to every image produced from a prompt; its significance lies in the court’s attention to the actual creative process.
A different situation arose in Thaler v. Perlmutter. The applicant sought registration of an image generated by an AI system and named the system as its sole author. The Court of Appeals for the District of Columbia Circuit held that US copyright law requires a human author. The Supreme Court subsequently declined to review the case on 2 March 2026, leaving the appellate judgment in force [8]. That refusal to hear the appeal was not a separate decision on the merits. The dispute concerned an image attributed solely to an autonomous system, not a work in which a person’s creative contribution had been established.
Taken together, the cases suggest that the use of AI should not, in itself, determine the outcome of a copyright claim. What matters is whether a person’s contribution can be identified in the expressive features of the particular result. Pressing a button, paying for a subscription or owning the device used to access the model would not, without more, justify recognising that person as the author.
Paragraph 80 of Resolution No. 10 of the Plenum of the Supreme Court of the Russian Federation, dated 23 April 2019, provides an important point of reference. The Plenum makes clear that a work may be created through technical means; the use of such means does not remove its creative character. Conversely, an image produced by technical equipment without human creative activity is not protected by copyright. The Resolution also establishes a presumption of creative character unless the contrary is proved. In the context of generative AI, this presumption does not remove the need to identify the person whose creative activity is embodied in the final expression [9].
This clarification is relevant to AI-assisted works because the use of technology cannot, by itself, serve as a reason to refuse copyright protection. The example of an automatically operating surveillance camera illustrates the other side of the issue: producing an image is not enough if no human creative activity shaped it. The court must therefore consider both the general presumption of creative character and the statutory requirement that an author be a citizen under Articles 1228 and 1257 of the Civil Code.
Who Is the Author and Who Owns the Exclusive Right?
Authorship and ownership of the exclusive right are connected, but they answer different legal questions. Under paragraph 3 of Article 1228 of the Civil Code, the exclusive right initially belongs to the person whose creative work produced the result. It may subsequently be transferred by agreement or pass to another person on a statutory basis. Article 1229, in turn, allows the holder of an exclusive right to be either an individual or a legal entity. The analysis must therefore identify the human author first and then establish who holds the exclusive right under the applicable law or contract.
This distinction is particularly important in employment relations. Under Article 1295 of the Civil Code, the employee remains the author of a service work created within the limits of their employment duties, while the exclusive right normally belongs to the employer, unless the employment or civil law contract provides otherwise. This rule applies only where the work was created within the employee’s duties; the mere fact that the employee used the employer’s equipment or materials is not sufficient. If, within three years after the work is placed at the employer’s disposal, the employer neither begins to use it, transfers the exclusive right to another person, nor informs the author that the work is to be kept secret, the exclusive right returns to the author. If the employer takes any of these steps within that three-year period, the author is entitled to remuneration. Thus, in an AI-assisted project, Article 1295 may determine the economic right holder, but only after a human employee has been identified as the author of the relevant work.
The Civil Code also contains separate default rules for commissioned works and works created while performing certain contracts. Under Article 1296, where a contract specifically requires the creation of a work, the exclusive right generally belongs to the customer unless the parties agree otherwise. If the customer holds that right, the contractor may use the work for its own needs under a free, non-exclusive licence, unless the contract provides otherwise; an author who does not hold the exclusive right is entitled to remuneration. Article 1297 applies where a work is created while performing a works contract or a contract for research, development or technological work that did not expressly require its creation. In that situation, the exclusive right generally belongs to the contractor, while the customer may use the work for the purposes of the contract under a non-exclusive licence without additional payment, unless the contract provides otherwise. These provisions show why payment for an AI service or a commissioned creative task does not, by itself, answer the question of ownership. The legal nature of the contract, the identity of the human author and the terms agreed by the parties must be examined together.
A user may be recognised as the author of an AI-assisted result if their creative decisions are apparent in its final form. The position of a general-purpose model developer is usually different: creating the tool does not necessarily involve deciding the expressive features of each output later produced with it. The same reasoning applies to a platform operator whose role is simply to provide access to the model. This does not affect any separate rights that may exist in the software, model or interface; those rights should not be confused with copyright in a particular generated result.
Proving the nature and extent of a user’s contribution may be difficult in practice. Prompt histories, earlier versions, parameter settings, source files, editing records, drafts and correspondence may help to reconstruct the process. None of these materials is conclusive on its own: the evidence must be considered alongside the expressive features of the final work. Paragraphs 109 and 110 of Resolution No. 10 are relevant here. They address the presumption arising where a person is identified as the author on a work and the use of other evidence when authorship is disputed. Keeping records of the generation and editing process may therefore prove useful in future disputes.
Service terms and platform agreements may regulate access to generated results, their commercial use, licences and the contractual allocation of economic interests. However, such terms cannot by themselves turn a person who made no creative contribution into the author under Russian copyright law. Contractual rights and the personal status of the author must be considered separately.
Comparative legal approaches
The United Kingdom has a special statutory rule for computer-generated works. Section 9(3) of the Copyright, Designs and Patents Act 1988 provides that, for a literary, dramatic, musical or artistic work which is computer-generated, the author is deemed to be the person by whom the arrangements necessary for the creation of the work are undertaken. Section 178 defines a computer-generated work as one generated by computer in circumstances in which there is no human author. Under section 12(7), copyright in such a work lasts for 50 years from the end of the calendar year in which it was made [10].
The UK rule offers a way of allocating rights where a work has no human author in the ordinary sense. The UK Government’s 2026 report notes that, where a general-purpose AI system generates an output in response to a user prompt, the person who entered the prompt will usually be treated as the person who made the arrangements necessary for its creation. This may provide a starting point, but it does not resolve every case involving several participants in the process. [11]
The same report identifies an uncertainty in the current framework. Section 9(3) remains in force, but its application is difficult to reconcile with the originality requirement for works created without a human author. The Government’s preferred option was to remove the specific protection for computer-generated works unless consultation responses provided sufficient evidence of its positive effects. This is a policy proposal, not an enacted amendment: sections 9(3), 12(7) and 178 remain in force unless Parliament changes them.
The United States follows a human-authorship approach. In Thaler v. Perlmutter, the Court of Appeals for the District of Columbia Circuit refused to recognise an AI system as the author of an autonomously generated image. The US Copyright Office’s 2025 report clarifies that the use of AI does not exclude protection where human authorship can be identified in the final output, but prompts alone are not necessarily sufficient.
The Chinese decision in case (2023) Jing 0491 Min Chu No. 11279 is more fact-specific. The court did not recognise AI as an author. Instead, it examined the user’s creative contribution and found that the user’s choices and revisions were reflected in the final image.
The three jurisdictions use different legal techniques. UK legislation currently provides a rule for works without a human author, although the Government has questioned whether that rule should be retained. US law places the emphasis on human authorship, while the Beijing decision shows how a court may examine the user’s actual contribution in a particular dispute. The latter two approaches sit more comfortably with Articles 1228 and 1257 of the Russian Civil Code. The UK model is nevertheless relevant when considering whether fully autonomous outputs should be subject to a separate legal regime.
Fully autonomous AI-generated results and a possible special legal regime
Where no sufficient human creative contribution can be identified, the result may fall outside copyright protection under the current Russian model. That conclusion does not mean the output has no economic value. Rather, it raises the question whether the traditional concept of authorship is an appropriate basis for allocating legal protection in such circumstances.
One option would be to grant the exclusive right to the developer of the AI system. This would, however, be difficult to justify in every case. A developer of a general-purpose model may have no knowledge of the particular output, the person who generated it or the purpose for which it was created. Giving developers rights in all outputs could concentrate control over a wide range of results and restrict users. It would also risk confusing rights in the model itself with rights in individual outputs.
A different option is to grant the right to the user who initiated the generation [12;13]. Although straightforward, this approach would place a person who entered a short, general prompt in the same position as someone who organised a substantial project to produce a particular result. Chernyshov, for example, discusses a one-year exclusive right for the end user who formulated the prompt. That proposal is worth considering, but prompt entry alone should not automatically settle the question of entitlement.
I consider a separate, limited sui generis right for fully autonomous outputs to be one possible direction for Russian law. This approach takes account of Kalyatin’s discussion of the different participants in AI-assisted creation, including those performing organisational and producer functions [14]. It also engages with Chernyshov’s proposal for a one-year right in favour of the end user. The model suggested here is narrower in one important respect: a prompt, by itself, would not determine who receives protection. The proposed right would not be copyright and would confer neither authorship nor personal non-property rights. Its purpose would be to protect a specific organisational or investment interest connected with an output. The mere development of a general-purpose model would not give its developer such a right.
The law would also need a workable test for identifying the person entitled to this protection. One possibility is to focus on substantial organisational or financial responsibility for the particular generation project. Relevant circumstances might include initiating and organising the project, commissioning or financing the generation, bearing its costs or commercial risk, and being the first to fix or lawfully disclose the output. Creative choices should not form part of this test: where those choices shape the expression, the ordinary rules of copyright may apply instead. Routine prompt entry, an ordinary subscription fee or general development of the model would not be enough. If several persons share the relevant organisational functions, their respective interests could be settled by agreement, subject to the protection of any human author.
Any such protection should be carefully limited. It might prevent unauthorised commercial copying or distribution of the particular fixed output for a short period, such as one year. It should not extend to ideas, facts, styles, methods, the underlying model or training materials. Nor should it prevent another person from independently generating a similar result. A narrowly framed right of this kind would address certain organisational and economic interests without treating them as creative authorship.
The suggested one-year term is a policy choice, not a period established by current Russian law. Its purpose would be to reflect the limited contribution involved in fully autonomous generation and to avoid a long-term monopoly over an output with no human author. There is, however, no empirical basis here for treating one year as the only appropriate duration. The period would require further economic and legal assessment before any legislative change. The regime should also include suitable exceptions for private and non-commercial use, quotation, research and independent creation. A claimant would have to prove when the output was fixed, why they are entitled to the right and that the particular output was copied or distributed commercially without authorisation.
Recent Russian legislation and its significance
Federal Law No. 243-FZ of 26 July 2026, “On Support for the Development of Artificial Intelligence Technologies in the Russian Federation”, is relevant to this discussion. Article 10 requires persons providing access to large foundational AI models to inform users about the rights in results obtained through those models and the conditions governing access to, use of and storage of those results. Under Article 13, this requirement is due to enter into force on 1 March 2027.
The importance of Article 10 lies primarily in the information duty it places on providers. It does not determine who authored a particular result, recognise AI as a copyright subject or establish a general rule that rights belong to either the user or the developer. The provision therefore does not replace the rules in Part IV of the Civil Code, which remain central to identifying the author and the holder of the exclusive right.
Article 10 also addresses the use of information contained in copyrighted works for certain forms of computational analysis and training of sovereign or national large foundational AI models, subject to the statutory conditions, including lawful access to the work or its availability for analysis without technical restrictions. This provision concerns the permitted use of source materials in the development of specified models; it does not determine the authorship or ownership of each output subsequently generated by those models.
The new law is best understood as a measure aimed at making the use of large foundational models more transparent, rather than as a complete answer to questions of authorship. Its practical effect will depend on the information providers give users and on how their contractual terms operate alongside copyright rules and the actual creative contribution made in a particular case.
- Results and discussion
The analysis leads to several findings. First, under the current Russian framework, an AI system cannot be recognised as an author: Articles 1228 and 1257 of the Civil Code reserve authorship for a citizen whose creative work produces the result. The use of AI does not, however, rule out copyright protection. The relevant question is whether a person’s creative contribution can be seen in the expressive form of the finished work.
Second, authorship and ownership of the exclusive right must be distinguished. The exclusive right initially arises with the author under Article 1228, but it may be transferred or belong to another person under statutory rules. Article 1295 is particularly important for AI-assisted works created by employees: the employee remains the author, while the employer may hold the exclusive right.
Third, the roles of the user, developer and service provider must also be assessed separately. Their legal position depends on their actual involvement in producing the particular result. Federal Law No. 243-FZ adds an information duty for providers of large foundational models, but it does not settle authorship or establish a default holder of the exclusive right.
Finally, fully autonomous outputs require separate consideration. The absence of a human author should not be resolved by granting authorship to an AI system or automatically assigning copyright to the developer or user. A narrow sui generis right may be considered, provided that its subject, holder, duration and limits are clearly defined.
Conclusion
Generative AI puts pressure on familiar copyright concepts, but it does not make the question of human authorship disappear. A single output may involve several participants, whose contributions have very different legal significance. The fact that an AI system produces the immediate result does not, on its own, identify either the author or the holder of the exclusive right.
Under the current provisions of the Civil Code of the Russian Federation, an AI system should be treated as a tool rather than as an author. Copyright protection may arise where a person’s creative decisions are reflected in the expressive form of the final result. The mere use of a prompt, payment for access to a service or ownership of the technology should not be sufficient on its own.
The author of a work and the holder of the exclusive right need not be the same person. This follows from the general rules governing the transfer of exclusive rights and from the special provisions on works created in the course of employment. The legal analysis should therefore establish whether a human author exists before turning to the question of who holds the economic right under the relevant statutory or contractual rules.
The contractual analysis must be conducted only after the legal nature of the relationship has been identified: an author’s order under Article 1288, a commissioned work under Article 1296, a work created in the course of performing another contract under Article 1297, or a service work under Article 1295. These legal regimes have different default rules and cannot be treated as interchangeable.
For fully autonomous outputs, Russian law could consider a separate and limited economic right. Such a right should remain distinct from copyright and should not confer legal personality on an AI system. A balanced approach would protect works containing sufficient human creative contribution under ordinary copyright, apply clear rules to the allocation of exclusive rights, and consider a narrowly defined regime for autonomous outputs if the legislature finds it necessary.
The proposed sui generis model is a recommendation for legislative discussion, not a description of the law currently in force. Its adoption would require a separate assessment of its compatibility with existing copyright, competition and international intellectual property obligations.
References
1. Federal Law No. 243-FZ of 26 July 2026 “On Support for the Development of Artificial Intelligence Technologies in the Russian Federation”. Articles 10 and 13. Available at: https://publication.pravo.gov.ru/document/0001202607260003 (accessed: 11.11.2026).
2. Civil Code of the Russian Federation (Part Four) No. 230-FZ of 18 December 2006 (as amended). Articles 1228, 1229, 1257, 1259, 1288, 1295, 1296 and 1297.
3. Kalyatin, V.O. (2022). Establishing of Subject of Rights to Intellectual Property Created with Use of Artificial Intelligence [in Russian]. Law. Journal of the Higher School of Economics, 15(4), 24–50. DOI: 10.17323/2072-8166.2022.4.24.50.
4. Chernyshov, M.S. (2024). The Problem of Distribution of Intellectual Rights on Artificial Intelligence Generated Works [in Russian]. Works on Intellectual Property, 49(2), 124–137. DOI: 10.17323/tis.2024.21721.
5. Utkin, V. (2024). Problems of Copyright for the Products Created by Artificial Intelligence [in Russian]. Society and Economics, 2, 79–83. DOI: 10.31857/S0207367624020058.
6. United States Copyright Office. (2025). Copyright and Artificial Intelligence. Part 2: Copyrightability. Washington, D.C., 29 January 2025. Available at: https://www.copyright.gov/ai/ (accessed: 11.11.2026).
7. Beijing Internet Court. Civil Judgment (2023) Jing 0491 Min Chu No. 11279, 27 November 2023. Available at: https://english.bjinternetcourt.gov.cn/2023-12/28/c_688.htm (accessed: 11.11.2026).
8. Thaler v. Perlmutter, No. 23-5233, 130 F.4th 1039 (D.C. Cir. 2025), decided 18 March 2025; certiorari denied, No. 25-449 (U.S. Supreme Court, 2 March 2026).
9. Plenum of the Supreme Court of the Russian Federation. Resolution No. 10 of 23 April 2019 “On the Application of Part Four of the Civil Code of the Russian Federation”, paras. 80, 109–110. Available at: https://www.vsrf.ru/documents/all/27773 (accessed: 11.11.2026).
10. Copyright, Designs and Patents Act 1988 (United Kingdom). Sections 9(3), 12(7) and 178. Available at: https://www.legislation.gov.uk/ukpga/1988/48/contents (accessed: 11.11.2026).
11. UK Government. (2026). Report on Copyright and Artificial Intelligence. Published 18 March 2026, Section I: Computer-generated works. Available at: https://www.gov.uk/government/publications/report-and-impact-assessment-on-copyright-and-artificial-intelligence/report-on-copyright-and-artificial-intelligence (accessed: 11.11.2026).
12. Privalov, N.G. (2025). Institutionalization of artificial intelligence: interdisciplinary aspect//Journal of Political Studies. Volume 9. № 4. P.3-27. DOI https://doi.org/10.12737/2587-6295-2025-9-4-3-27
13. Privalova, N.I. (2026). Homo traditum, homo technicus and artificial intelligence: spiritual aspects of interaction//Research. Social and Humanitarian Research and Technology, № 1 (54). P.68-80.
14. Kalyatin, V.O. (2026). On Certain Aspects of Defining the Subject of Intellectual Property Created with the Use of Artificial Intelligence [in Russian]. Journal of the Court for Intellectual Property Rights, 1(51), 17–24. DOI: 10.58741/23134852_2026_1_2.
