Introduction

Copyright law has traditionally assumed a relatively intelligible relationship between the creator and the work. Technology may facilitate writing, composition, photography, or illustration, but the creative choices can ordinarily be traced to the person who made them. Generative artificial intelligence changes that relationship because a system can now supply substantial elements of expression after receiving instructions that range from detailed creative direction to a short and general prompt. The legal question is therefore not simply whether technology was used, but whether the final expression can still be meaningfully attributed to a person whose intellectual and creative choices produced the work. 

The relevant legal framework is Law No. 28 of 2014 on Copyright ("Copyright Law"), which places human intellectual activity at the centre of protection. Article 1 point 2 of the Copyright Law defines a Creator as one or more persons who individually or jointly produce a work that is distinctive and personal. Article 1 point 3 of the Copyright Law defines a Work as an expression in science, art or literature produced through inspiration, ability, thought, imagination, dexterity, skill or expertise and expressed in a tangible form. Article 1, point 27 of the Copyright Law defines a Person as a natural person or legal entity. An AI system does not fall within that statutory category. The current framework therefore directs the inquiry away from the machine as legal author and toward the person whose intellectual choices can be connected to the final expression. AI-assisted creation nevertheless exists on a spectrum, where AI may implement a concept substantially designed and refined by a human, or a user may give a general instruction and accept the resulting output with little intervention. The relevant inquiry is the quality of the human contribution, not the mere fact that AI was used. Preserving and evidencing meaningful human creative control therefore becomes central to maintaining a defensible basis for copyright protection in AI-assisted works. 

Copyright, Authorship and Ownership under the Existing Framework 

The first distinction is between authorship and ownership. Article 1 point 4 of the Copyright Law provides that a Copyright Holder may be the Creator as the owner of copyright, a party that lawfully receives the right from the Creator, or a further lawful transferee. The person who qualifies as Creator and the person who ultimately holds the economic rights are therefore not always the same. Article 31 of the Copyright Law also establishes a rebuttable presumption in favour of the person identified as the Creator through the forms of attribution recognised by that provision, including identification on the work or in the copyright record. That presumption does not replace the substantive inquiry into authorship, but it adds an evidentiary layer to the question of who may initially assert the copyright. Article 37 of the Copyright Law adds a further statutory presumption: unless proven otherwise, a legal entity that announces, distributes, or communicates a work originating from it without naming an individual Creator is deemed to be the Creator. That attribution rule does not, however, make a legal entity the Creator merely because it owns or deploys an AI system; the output must still satisfy the statutory requirements for a protected Work. In an AI-assisted process, these provisions matter because the user of the system, the person who develops the creative concept, an employer or commissioning party, a legal entity releasing the work, and the provider of the AI system may all be different actors. The analysis should first identify the statutory basis on which authorship may be attributed and only then determine how the resulting rights are held or transferred. 

The attribution provisions of the Copyright Law are useful because they show that design and execution need not always rest in the same hands. Article 33 of the Copyright Law addresses works consisting of separate parts produced by two or more persons, while Article 34 of the Copyright Law provides that, where a work is designed by one person and realised by another under the leadership and supervision of the designer, the designer is regarded as the Creator. The official explanation to Article 34 of the Copyright Law describes such leadership and supervision in terms of guidance, direction or correction. Article 34 should not be mechanically extended to generative AI because it contemplates another Person carrying out the work, whereas an AI system is not a Person under Article 1 point 27 of the Copyright Law. Its structure is nevertheless instructive: authorship may follow the source of creative design and meaningful direction rather than the actor performing every technical step. The three elements identified in the official explanation to Article 34 — guidance, direction, and correction — offer a useful, if imperfect, template for distinguishing meaningful human control from passive acceptance of AI output. 

Ownership requires a separate examination. Article 36 of the Copyright Law regulates works made in an employment relationship or on commission, while Article 16 of the Copyright Law permits copyright to be transferred, in whole or in part, through the lawful causes identified in that provision, including a written agreement. These rules make contractual allocation relevant where employees, freelancers, agencies or commissioned creators use generative AI. Payment for a work or ownership of the technology used to create it does not, by itself, resolve authorship. The same applies to an AI provider: ownership of the model, software or platform does not automatically make the provider the Creator of every output. Provider terms may allocate contractual rights in inputs and outputs, but they cannot replace the statutory inquiry into whether copyright exists and who qualifies as Creator or Copyright Holder under the Copyright Law. 

Human Creative Contribution: Its Place in AI-Assisted Works 

Human creative contribution should therefore operate as the central interpretive concept. Copyright Law does n ot prescribe a numerical threshold of creativity or specify how many prompts, edits or iterations are required before an AI-assisted work is protected. The better approach is qualitative: whether the final expression embodies identifiable human choices of the kind contemplated by Article 1 point 3 of the Copyright Law, including thought, imagination, skill and expertise. 

Creative conception is relevant, but an idea alone does not settle authorship. Article 41 letter b of the Copyright Law excludes ideas, procedures, systems, methods, concepts, principles, discoveries and data from copyright protection even where they have been disclosed or described in a work. Prompting therefore matters only to the extent that it demonstrates direction over protected expression rather than merely stating a desired subject, concept or style. The length or number of prompts should not become a formal test. A detailed instruction may still leave decisive expressive choices to the system, while a shorter prompt may form part of a broader human process involving sketches, sequencing, composition, editing or other creative decisions. The question is whether the human choices can be traced into the expression for which protection is claimed. 

Selection, curation and arrangement may also evidence authorship where they involve aesthetic judgment rather than technical quality control. A user may reject outputs, combine elements, determine sequence or placement, and select alternatives because they serve a larger creative objective. Editing and modification may make the connection clearer through rewriting, redrawing, recomposition, or replacement of generated components. Conversely, a person who enters a general prompt and accepts an output substantially as generated may have difficulty showing that the distinctive and personal features of the final work are attributable to that person's own intellectual contribution. The conclusion remains fact-specific: neither the AI user nor the AI provider becomes the Creator merely by operating or owning the system. Authorship should follow the identifiable human creative process reflected in the final work. 

From Human Creative Contribution to Copyright Risk Management 

The practical consequence is that risk mitigation should preserve and evidence the human creative process. This does not mean that documentation creates copyright. Article 1 point 1 of the Copyright Law provides that copyright arises automatically on a declarative basis once a work is expressed in tangible form, subject to the statutory limitations. Evidence serves a different function: it helps demonstrate how the final expression was produced, who made the relevant creative decisions, and how the resulting rights were allocated. 

For material AI-assisted projects, contemporaneous records should show the human conception and direction behind the work. An outline, storyboard, sketch, musical structure, design brief, or draft can establish the intended creative result. Relevant prompts, generated versions, selections, combinations, and editing stages can then show how the final form emerged. The purpose of preserving these materials is not to prove that AI was used, but to identify which expressive choices remained attributable to the human creator. Meaningful human review should also occur before finalisation; generation should not automatically be treated as the last creative act where the objective is to maintain a strong basis for copyright protection. 

Contractual arrangements should align ownership with that creative process. Where employees, freelancers, agencies or commissioned creators are involved, agreements should identify relevant creative roles and allocate economic rights consistently with Article 16 and Article 36 of the Copyright Law. The applicable AI-provider terms should also be reviewed to determine whether the provider claims rights in inputs or outputs or imposes restrictions affecting commercial use. A contract cannot create copyright where the statutory basis for human authorship is absent, but unclear contractual arrangements can reduce the commercial value of a work even where the human contribution is strong.

Copyright recordal may provide an additional evidentiary layer but should not substitute for the underlying authorship analysis. Article 31 of the Copyright Law may support a rebuttable presumption of authorship where a person is identified through the forms recognised by that provision. Separately, Article 69 paragraph (4) of the Copyright Law provides that, unless proven otherwise, a certificate of recordal constitutes initial evidence of ownership of a Work or Related Rights product. Article 72 of the Copyright Law further provides that recordal in the general register does not constitute endorsement of the content, meaning, purpose or form of the recorded work. Recordal is therefore most useful when supported by contemporaneous material showing the human creative process and by a coherent contractual chain explaining how the resulting rights are held. 

The level of mitigation should remain proportionate to the value and exposure of the work. Where an AI-assisted work is intended to become a material intellectual property asset, preserving the creation history and clarifying the chain of rights can help expand creative capacity without weakening the legal basis of the resulting copyright claim. 

Conclusion 

Generative AI does not make the existing copyright framework irrelevant; it makes the distinction between human creativity and technological execution more important. Copyright Law directs protection toward works that are distinctive and personal and arise from human intellectual capacities. Its attribution and ownership provisions provide a workable basis for asking who designed, directed, and shaped the final expression and who lawfully holds the resulting rights. Ownership of an AI system does not automatically confer authorship over its outputs, while use of the system does not automatically make the user a Creator. 

Risk mitigation should preserve the legal basis for copyright protection, not manufacture it. Meaningful human control must exist in fact, and contemporaneous evidence should connect that control to the final expression. Human creative control is therefore not merely an abstract authorship principle; it is the central risk-mitigation mechanism for preserving copyright protection in AI-assisted works. 

These risk-mitigation measures — identifying the relevant human creative roles, aligning employment or commissioning arrangements with the Copyright Law, reviewing AI provider terms, and structuring proportionate evidence preservation — can be supported by the role of legal counsel. Legal counsel can help map which actors hold which creative role across an AI-assisted workflow and align the underlying agreements with the Copyright Law. Legal counsel can also assist in reviewing AI-provider terms and structuring an evidence-preservation practice proportionate to the value of the work, so that the legal position claimed for a work is supported by the process through which it was actually created and acquired.