From Algorithms to Authorship: The Legal Conundrum of Assigning Copyright in AI-Generated Creations

Aditya Singh is a Third-year law student at RGNUL, Punjab, pursuing legal research and writing.

Introduction

The subject of who owns ‘creations’ created by artificial intelligence is a complex and evolving area of intellectual property law that raises fundamental questions about creativity, the nature of authorship, and the role of Generative AI in this creative process. At the heart of this debate is a single question: should AI-produced works be granted copyright protection, and if so, who is the entitled owner of these works? Many works have emerged because of the increasing integration of AI with creative endeavors, including art, music, literature, and even screenplays. These AI-generated results call into question the long-held belief that creativity is founded on human talent. Historically, the legal framework for copyright has been built on a concept of human authorship, which asserts that creative work is merely an application and extension into the material form of the human personality and cerebral investment that went into generating it. However, AI disrupts this paradigm by injecting a non-human element into the creative process, raising questions about whether present copyright laws apply to software-created works.

Current Copyright Landscape

Different jurisdictions treat AI and copyright differently. Copyright laws in countries such as the United Kingdom and India provide unique provisions for computer-generated works. Such rules typically grant copyright to whoever arranged the conditions essential for the creation of the work. However, this procedure is not universally acknowledged or followed, and there is no clear framework for the legal regulation of autonomous inventions in many jurisdictions.

An issue at the heart of the debate over copyright protection for works produced by AI is one of ownership. Three main models have been proposed: 

Software Developer as Owner: This model gives copyright to the person who creates the AI software, with the rationale that it is his skill and judgment that are reflected in the work created by AIs.  However, it could potentially marginalize users who actively shape the outputs through their inputs.

Software User as Owner: Under this model, the user of an AI system takes on copyright for their decision-making role. He or she organizes data input and provides guiding instructions in accordance with aesthetic goals, and it is these efforts that primarily shape the final work. This approach emphasizes the user's creative involvement but may lead to disputes over the relative contributions of the user and the developer, complicating ownership claims.

AI Owner as Owner: This approach attributes ownership of the AI-generated work to the owner of the AI system, treating the output as the natural result—or ‘fruit’—of owning and operating the system, simplifying ownership disputes, encourages investment in AI systems, and provides a clear legal framework for commercialization, however potential monopolization of AI generated content by large corporations can pose significant challenges.

Ambiguity comes into play in such provisions, evident in The Indian Copyright Act, 1957, where section 2(d)(vi) defines the author “in relation to any literary, dramatic, musical or artistic work that is computer-generated, the person who causes the work to be created”. Whether it is the developer of the AI or the person whose inputs led to the creation of the work, identifying the “person who caused the work to be created” becomes a matter of legal interpretation. This is because only natural individuals are currently recognized as authors by the law, and it is critical to define and interpret the term "person" in this context. The legal position of AI—whether it qualifies as a "person" under the law and, if so, to what extent—must be clarified by the legislature and the courts.

The originality of a work is one of the criteria used to establish whether it is protected by copyright. Section 13 of the Indian Copyright Act protects "original literary, dramatic, musical, and artistic works." The provision does not define originality clearly, instead leaving it up to the courts to determine if an AI-generated work is "original" enough, such as in Eastern Book Company v. D.B. Modak, where it was held that the creativity standard applied is not that something must be novel or non-obvious, but some amount of creativity in the work to claim a copyright is required, implying that the creativity resulting from the data inputs provided by the user and utilized by the AI can fall within the scope of copyright protection under the law.

Similar ambiguities exist in the laws of the UK and Ireland, where provisions for computer-generated works exist.

Current Judicial Perspectives

The issue became more prominent with the DABUS case where applications to patent inventions of Stephen Thaler's Al system DABUS were rejected in a number of jurisdictions, including Australia, the United Kingdom, the United States, New Zealand, and the European Patent Office, due to a common classification in their patent laws that prohibits the grant of a patent to an Al system as the inventor. However, the South African Companies and Intellectual Property Commission (CIPC) accepted Stephen's Patent Cooperation Treaty for a patent on DABUS technologies on June 24, 2021. In July 2021, CIPC issued a notice of issuance for the patent, and DABUS became the first AI system to be acknowledged as an inventor and granted a patent for its inventions. Although the DABUS case is primarily concerned with patent law inventorship, it also has significant ramifications for the Al and copyright issue as the case brought to the forefront the legal system's challenge in applying existing frameworks to non-human entities as creators or inventors, highlighting a similar gap in copyright law where authorship is traditionally reserved for natural persons.

An extension to this was seen in the RAGHAV case where India's copyright office acknowledged an AI system, RAGHAV, as a co-author of an artistic work and registered the application for copyright protection. The claim was earlier refused in the first instance by the copyright office after Ankit Sahni, the designer of the AI system RAGHAV, filed an application mentioning the AI system as the sole author of such work. Later, the copyright office filed a notice of revocation of such registration, claiming that it had been granted wrongly, and urged the human co-author, Mr. Sahni, to review the legal position of the AI system RAGHAV. The application is still "registered" on the copyright office's website, but the court has yet to rule.

However, in Tencent v Shanghai Ying Xun Technology Co. Ltd, Tencent argued successfully in a Chinese court that their software developer should be considered the owner of an article created by their AI software, Dreamwriter. Dreamwriter, designed to generate articles, wrote a piece on the Shanghai stock market by collecting and analyzing data, verifying it, and then writing and publishing the article. The defendant claimed the article lacked copyrightability due to the absence of human creativity. However, the court recognized human originality in Dreamwriter's development process. They noted that while Dreamwriter autonomously produced the article based on pre-established rules, algorithms, and templates, this operation reflected the developers' significant input. This input included selecting and arranging data types and formats, setting conditions for article writing, structuring templates, and training the verification algorithm. The court's perspective aligned with the originality doctrine in copyright law, akin to cases involving compilations, where the intellectual creation is seen in the choice or arrangement of existing content. The court concluded that originality was present in the developer's criteria for selecting and arranging data, which the AI then used to complete the content. This decision marked a notable stance in recognizing the role of AI developers in imparting originality to AI-generated works, providing a distinct perspective on the issue.

Way Forward

The rapid rise of generative AI has exposed significant gaps and inconsistencies in existing copyright frameworks, as current laws across various jurisdictions were designed with human authorship at their core. This creates uncertainty over how AI-generated works should be protected and who should hold the associated rights. The absence of a harmonized global approach complicates the protection, ownership, and commercialization of AI outputs, making it imperative to address these challenges through a coherent legal framework.

The AI-owner approach serves as a plausible solution to this issue, which, as mentioned earlier, posits that the owner of the AI is to be taken as the default owner of any IP rights for works generated by that system. The advocacy of this approach arises due to several reasons. Firstly, it provides a straightforward and efficient method to assign ownership, avoiding the need to evaluate the relative contributions of various participants in the creative process (such as users or AI developers). Secondly, it also ensures that the issue of ownership does not fall into the public domain, which risks undervaluing AI developers' efforts, discourages investment in AI research, and can lead to unchecked usage and potential ethical and quality concerns. The challenge becomes more pronounced in the other two approaches of assigning ownership to either the software user or software developer, as they can result in ambiguous and contested claims, with uncertainty over whether the developer's programming or the user's interaction with the AI constitutes the predominant creative force. Thirdly, aside from practical justifications, this approach finds support in the existing copyright statues, including but not limited to, Section 2(d)(vi) of Indian Copyright Act which defines an "author" for computer-generated works as the person who causes the work to be created, while Section 9(3) of UK Copyright, Designs and Patents Act 1988 states that copyright in a computer-generated work belongs to the person who makes the necessary arrangements for its creation, and AI owners aptly fit into such definition. Similarly, Section 201 (b) of the US Copyright Act states that employers or commissioning parties retain copyright in works created by employees or under contractual agreements. Similarly, AI owners could be seen as the commissioning party, given their ownership and operational control over the AI system. 

This approach also reduces the chances of complex co-ownership arrangements arising where both developers and users contribute considerably to the creation. Co-ownership usually results in a complicated management-rights situation, high transaction costs and costly legal battles in case of disputes, all of which can threaten the value of the work’s copyrights, as unresolved or disputed ownership can create legal uncertainty, making it difficult to license, sell, or enforce rights over the work. This ambiguity may deter potential investors or users who fear litigation or overlapping claims, thereby diminishing commercial viability and market confidence in the work. Therefore, this simplifies ownership and streamlines legal processes as well as commercial ones. It also provides owners with an incentive to widen the availability of their technology and develop it. For example, had IBM been unable to claim ownership of the output generated by its Watson AI—such as data-driven insights, reports, or content created through its cognitive computing capabilities—it would not have shared the technology. Flexible arrangements for ownership are possible: Users who are willing and able to pay a premium for exclusive rights can pay higher licensing fees or negotiate specific agreements between themselves and the owners of AI. Alternatively, developers of AI not requiring ownership over output could transfer rights to users through specific contractual arrangements, which may outline terms such as ownership assignment, scope of use, duration, or royalty-based licensing. These contracts would enable flexible commercial setups—such as providing exclusive rights to large enterprises, granting limited rights for campaign-based content, or enabling open access under “freemium” models—depending on the user’s needs and bargaining position. This would accommodate a wide range of commercial requirements, including exclusivity, time-bound usage, and sector-specific rights, while supporting investment recovery strategies like upfront licensing fees, periodic royalty models,orstrategicpartnershipsfortechnologydissemination.There are certain challenges to this approach as well. Problems arise when applying this approach in practice, and especially so for an AI system licensed commercially. For example, if a media company contracts an AI system for creating content, it rightfully would expect ownership or exclusive usage of the content. Such a denial of rights could discourage use of generative AI itself, as shown in the Feilin case, where the Beijing Internet Court dealt with a dispute over a data report generated with assistance from an AI-powered legal database. Although the court held that the software itself could not be considered the author, it also declined to recognize the developer or owner of the AI tool as the rightful holder of the work. Instead, it emphasized the role of the user, who customized data inputs, filtered results, and structured the final expression as central to the creative process. The court reasoned that the user, having tailored the output to their specific needs and invested in its dissemination, held the more legitimate claim to any communicative and economic value arising from the content. For end users with no room to negotiate, however, this problem is magnified. To address this challenge, a practical solution lies in adopting default ownership by AI owners with the flexibility to contractually assign or license rights to users. This model balances the developer’s investment and control with the commercial expectations of users, particularly in industries like media, where content ownership is critical. Clear contractual mechanisms can thus ensure mutual benefit, legal certainty, and broader adoption of generative AI across sectors.

In conclusion, the issue surrounding ownership of AI-generated works remains ambiguous and requires urgent legal clarity, especially with AI becoming a significant part of modern industries. A clear and consistent global framework is essential to address these challenges. The AI-owner approach offers an efficient way to assign ownership while encouraging innovation. However, practical and ethical challenges highlight the importance of more clearer reforms, fulfilling the need for harmonized laws that will protect intellectual property rights and promote trust and investment in generative AI technologies.