Authorship Beyond the Material Object
Exhibitions have traditionally been based on the presence of the artwork as a concrete object: a painting, sculpture, photograph, or other material artefact that can be preserved, identified, and presented to the public. In contemporary museum studies, however, exhibitions are increasingly understood not simply as arrangements of objects, but as cultural and communicative systems that organise relationships between the artwork, the space, and the visitor (Hooper-Greenhill, 2000). The meaning of an artwork does not arise solely from its material characteristics, but also from the context in which it is presented.
Throughout the twentieth century, the very concept of the “artwork” gradually began to move away from the idea of an autonomous object. Installation and so-called site-specific art turned space itself into part of the artistic act, while video art, interactive and multimedia forms introduced sound, movement, light, software, and audience participation (Bishop, 2005). In this process, not only did the ways in which artworks were exhibited change, but so too did the question of what can actually be classified as an artwork.
Digital technologies have brought this transformation to a new stage. An artwork can now take the form of a 3d model, be reconstructed from archival data, be spatially simulated through projection or a virtual environment, respond to the actions of a visitor, or be generated by an algorithmic system. In immersive art, the boundary between artwork and space is becoming increasingly blurred, while in generative art the form of the artwork itself may not be predetermined by the artist.
This gives rise to a problem that cannot be resolved simply by asking whether artificial intelligence is an “instrument” or a “creator”. If an artwork can be reconstructed, simulated, reproduced, or generated in multiple forms, what makes a particular realisation an authentic artwork? More importantly, if the final result is merely one of the possible realisations of a given system, where does authorship begin and end?
This is precisely the question at the centre of this study: what happens to the concept of the artwork and to authorship when an artwork is the product of digital technologies and can be reconstructed, spatially simulated, and generated anew?
A New Concept of Authorship
This question did not begin with artificial intelligence. As early as the second half of the twentieth century, art theory began to challenge the seemingly self-evident relationship between the artwork and its author. In The Death of the Author (1967), Roland Barthes questioned the idea that the meaning of a work can be reduced to the intentions of its creator. Michel Foucault (1969), in turn, conceptualised the author as an “author function” — a cultural and institutional mechanism through which works are classified, organised, and assigned a particular status.
These ideas acquire new significance in the digital environment. In the case of a traditional artwork, the physical object makes it relatively easy to trace the relationship between the act of creation, the material result, and the author. With digital artworks, this relationship becomes more complex. The result may be the outcome of a series of successive decisions: concept development, data selection, choice of software or model, setting of parameters, generation, selection, editing, and spatial presentation.
This means that authorship is no longer necessarily concentrated in a single moment of creation. It can instead be distributed across a series of creative decisions.
The problem becomes particularly visible in generative art. An artist may establish a system of rules, parameters, or instructions without controlling the specific visual form that the system will produce. Research into early computer art has already demonstrated that algorithmically generated forms challenge traditional understandings of intention and control (McCormack, Gifford & Hutchings, 2019). With generative AI, this relationship is even more complex, as the result may emerge through an interaction between human conception, a trained model, and a particular generation procedure.
This leads to an important distinction: creating an artwork and creating the conditions under which an artwork can emerge are no longer necessarily the same act.
This is where the contemporary problem of authorship lies.
This does not mean that the algorithm automatically acquires the status of an author. On the contrary, it may be precisely the human decisions that give the entire process its authorial structure. The difference is that this structure is no longer necessarily located in the final visual object.
This logic is also reflected beyond art theory. In the current practice of the U.S. Copyright Office, for example, the use of AI does not in itself preclude human authorship, but merely entering a prompt is not considered sufficient. What matters are the human creative decisions involved in selecting, arranging, and modifying the generated material (U.S. Copyright Office, 2025).
The question, therefore, is not simply whether an artwork was created using AI, but what role the human played in the process through which it came into being.
Not the End of Authorship, but a Change in Its Form
This perspective also allows us to reconsider the broader fear that digital technologies and generative AI will “eliminate” the author. The more likely outcome is the opposite: authorship will become more complex, more distributed, and increasingly dependent on the ability to design and manage creative processes.
This does not mean that every use of AI constitutes artistic authorship. Nor does it mean that the algorithm should be romanticised as a new creative subject. On the contrary, this is precisely why it is necessary to distinguish between automated production, human selection, conceptual design, and artistic decision-making.
This is also where the need for cultural and artistic institutions to adapt becomes apparent. Digital and generative art are not a temporary technological trend that the cultural sector can simply wait out. They are part of a broader transformation in the ways cultural works are created, presented, and perceived. The task, therefore, is not to preserve art as it existed before digitalisation, but to develop adequate criteria for evaluating, preserving, and interpreting works whose characteristics have already changed.
The Japanese approach is particularly illustrative of this transformation. Japan’s Agency for Cultural Affairs (Government of Japan) takes the position that material autonomously generated by AI does not, in principle, constitute a copyright-protected work, whereas an AI-generated result may acquire the status of a work when a human uses AI as a tool for their own creative expression. In making this assessment, both creative intent and human creative contribution are taken into account, with explicit emphasis on evaluating the entire process of using AI rather than merely the final result. This is particularly close to the argument advanced in this article concerning authorship as a process: institutional attention should shift from the question of what was generated to how it was generated and what the human actually did within that process.
In the United States, the institutional response is more strongly centred on the question of human authorship. In its report Copyright and Artificial Intelligence, Part 2: Copyrightability (2025), the U.S. Copyright Office concludes that the use of AI as a tool does not, in itself, preclude copyright protection. What matters is the nature and degree of human creative contribution: a work may be protected when the human author determines a sufficient amount of its expressive elements, including through creative selection, coordination, arrangement, or modification of AI-generated material. At the same time, the U.S. Copyright Office explicitly concludes that merely entering a prompt is not sufficient where the prompt does not provide adequate control over the expressive characteristics of the resulting output (U.S. Copyright Office, 2025).
Particularly significant is the fact that the American approach does not establish a separate legal category for “AI art”. Instead, it applies the existing criterion of human authorship to a new technological environment. The report emphasises that a work created with the assistance of AI may be protected where AI assists rather than replaces human creative activity. Conversely, purely AI-generated material, or material over whose expressive elements the human has not exercised sufficient control, does not qualify for copyright protection (U.S. Copyright Office, 2025).
In the European context, the question of authorship of AI-generated works remains unresolved, and the approach is somewhat different. In Copyright of AI-generated works: Approaches in the EU and beyond (2025), the European Parliament notes that EU law currently contains no specific Union-wide rules determining when a result created with the assistance of AI constitutes a copyright-protected work, nor what degree of AI involvement is compatible with human authorship. Instead, the general principle of European copyright law applies: a work must be original and constitute the author’s “own intellectual creation”, resulting from the author’s free and creative choices (European Parliament, 2025).
It is equally significant that the AI Act does not attempt to resolve this question. The Regulation introduces transparency requirements concerning AI-generated content, including artistic content, but does not recognise AI as an author or establish criteria for copyright protection of its outputs. From 2 August 2026, certain AI-generated or manipulated audio, image, video, and text outputs must be identifiable as artificially generated or manipulated. European regulation thus makes an important distinction between the origin of content and its status as an authored work: the fact that a work has been generated using AI does not, in itself, answer the question of whether it constitutes a work under copyright law.
This is where an interesting paradox emerges for the present study. The European legal framework continues to search for a human creative trace, while the creative process itself is increasingly distributed among humans, data, models, algorithms, and subsequent acts of selection. The question, therefore, is no longer simply whether the final result “looks like art”, but whether and where in the process there are sufficient free and creative human choices for the result to be attributed to an author. This gap between the traditional concept of human authorship and the actual process of AI-assisted creation demonstrates the need to conceptualise authorship not only through the final product, but also through its provenance and process of creation.
From Comparison to a New Institutional Model
A comparison of the three approaches shows that the institutional response to AI in the arts is still in the process of formation. Japan places emphasis on creative intent and contribution throughout the entire process; the United States focuses on the degree of human control over expressive form; while the European Union retains the broader and more conservative criterion of “own intellectual creation”, without yet having established a specific regime for AI-generated works. These differences are not merely technical matters of law. They demonstrate that institutions are still attempting to answer a new question using categories developed for works whose origins could be relatively easily traced to a particular human being and a particular material result.
This is precisely where the central problem lies: AI does not make authorship impossible; it makes the traditional way of establishing authorship insufficient.
This is not a call to abandon traditional categories. On the contrary, precisely because originality, authorship, and authenticity remain fundamental to cultural institutions, they must be expanded so that they can accommodate works that no longer function as singular material objects. Institutions should not have to choose between traditional art and AI, but should develop criteria capable of distinguishing different degrees and forms of human creativity within technologically assisted processes.
In this sense, the comparison between Japan, the United States, and Europe reveals not so much which system is “correct”, but rather that the world is still searching for an adequate institutional language for a new artistic reality. This search should not be interpreted as evidence that AI ought to be restricted or rejected. The absence of a final consensus is rather a sign that institutions are undergoing a process of adaptation. Adaptation, however, does not mean the uncritical acceptance of all AI-generated content. It means developing mechanisms capable of establishing human contribution, provenance, the degree of automation, and the history of creation.
And here the central thesis of this article becomes clear: digital transformation does not eliminate the author — it changes where we need to look for the author. Authorship can no longer be identified solely with the immediate production of the final image. It may lie with the person who designed the system, defined its rules, made the creative choices, and controlled the process through which the image became possible. In the age of generative AI, authorship is therefore increasingly less a characteristic of the product and more a characteristic of the process.
The museum is compelled to shift part of its attention from the object itself to the conditions that make it an artwork. Institutional adaptation should therefore not be understood as a retreat from traditional artistic values. On the contrary, it is necessary precisely so that these values can be applied to new forms of artistic work. What is needed is not a rejection of traditional concepts of the original, authorship, and authenticity, but an expansion of their scope. Otherwise, institutions risk turning technologically new works into old-style objects — preserving the image while losing the process; documenting the result while losing its provenance; identifying the author while concealing the network of human and technological decisions through which the work became possible.
This is precisely the institutional stake of digitalisation: the museum of the future will not simply preserve more digital objects; it will have to develop a different understanding of what it means to preserve an artwork. This shifts the institutional focus from preserving the object to preserving its identity.
And this is precisely why, instead of asking only “Who is the author?”, future institutions will also need to be able to answer another question: “How did this artwork come into being?” It is possible that this second question will gradually become a new criterion through which we recognise not only authorship, but also the authenticity of an artwork. Thus, digital transformation does not eliminate the author — it changes where we need to look for the author. Authorship gradually shifts from the product to the process, while authenticity shifts from the uniqueness of the object to the traceability of its provenance.
Bibliography:
- Agency for Cultural Affairs, Government of Japan. (2023). FY2023 Copyright Seminar “AI and Copyright”: Presentation Materials. Tokyo: Agency for Cultural Affairs, pp. 55–60. Official document
- Agency for Cultural Affairs, Government of Japan. (2024a). Concepts on AI and Copyright. Tokyo: Agency for Cultural Affairs, pp. 39–40. Official document
- Agency for Cultural Affairs, Government of Japan. (2024b). General Understanding on AI and Copyright in Japan – Overview. Tokyo: Agency for Cultural Affairs.
- Agency for Cultural Affairs, Government of Japan. (n.d.). Frequently Asked Questions. Tokyo: Agency for Cultural Affairs.
- Barthes, R. (1967). The Death of the Author. Aspen, 5–6.
- Bishop, C. (2005). Installation Art. London: Tate Publishing.
- European Parliament, European Parliamentary Research Service. (2025). Karttunen, S. Copyright of AI-generated works: Approaches in the EU and beyond. EPRS Members’ Research Service, Briefing, PE 782.585, December 2025. Official EPRS publication
- European Union. (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union.
- Foucault, M. (1969). What Is an Author? Lecture delivered at the Société française de philosophie, 22 February 1969. In D. F. Bouchard (Ed.), Language, Counter-Memory, Practice: Selected Essays and Interviews (pp. 113–138). Ithaca, NY: Cornell University Press, 1977.
- Hooper-Greenhill, E. (2000). Museums and the Interpretation of Visual Culture. London: Routledge.
- McCormack, J., Gifford, T., & Hutchings, P. (2019). Autonomy, authenticity, authorship and intention in computer-generated art. In A. Ekárt, A. Liapis, & M. L. Castro Pena (Eds.), Computational Intelligence in Music, Sound, Art and Design: 8th International Conference, EvoMUSART 2019 (pp. 35–50). Cham: Springer. https://doi.org/10.1007/978-3-030-16667-0_3
- U.S. Copyright Office. (2025). Copyright and Artificial Intelligence, Part 2: Copyrightability: A Report of the Register of Copyrights. Washington, D.C.: U.S. Copyright Office, January 2025. Official U.S. Copyright Office report
