Dr. Yatin Kathuria

Introduction

AI has undergone remarkable advancements, evolving from simple classification models to powerful generative models capable of producing creative outputs in numerous forms like text, images, audio, video content and more.  This unique capability of Generative AI made it the fastest-adopted technology ever.  OpenAI’s ChatGPT, for example, reached 100 million users within just two months of its launch, a milestone that took Facebook over four years to achieve.

While the potential of Generative AI models to augment human creativity is irrefutable, the growing frequency of AI-generated content is challenging the conventional understanding of copyright law. Courts, policymakers, and governments across various jurisdictions are currently grappling with the complex legal challenge of determining whether AI-generated content qualifies for protection under existing copyright frameworks or not.

Recently, several jurisdictions have started acknowledging the potential copyright protection of AI-generated works; however, their legal approaches diverge considerably. The US Copyright Office, through its Copyright and AI reports, has adopted a human-authorship-centred approach, recognising copyright protection for AI-assisted works where sufficient human creative expression is present.  The EU follows an “auxiliary tool” approach, protecting AI-assisted works where human users exercise creative control over the outputs generated by AI.  South Korea similarly recognises AI-assisted works where authorship can be attributed to a natural person, while Japan adopts a case-by-case assessment based on factors such as prompt complexity, iterations, output selection and human modifications. Whereas in India, the Suryast episode initially recognised RAGHAV AI and Ankit Sahni as co-authors, the subsequent withdrawal notice questioning AI’s status as an “author” under the Copyright Act, 1957, leaves the copyrightability and authorship of AI-generated works legally unsettled.  Ukraine, however, adopts a distinct sui generis regime, granting economic rights in autonomous AI-generated works that do not meet the originality threshold for copyright.  

The dearth of uniform international approaches to the copyrightability of AI-generated works shows substantial gaps in existing copyright jurisprudence. This blog therefore investigates two key questions: (a) Can AI-generated works satisfy the originality requirement under prevailing judicial standards? and (b) Who should be recognised as the legal owner—the AI itself or its user?  In answering these questions, the blog aims to contribute towards a more uniform legal approach to the copyright protection of AI-generated works across jurisdictions.

Can AI-generated works be deemed “original” under Copyright Laws?

Copyright law, through the concept of originality, tends to protect the uniqueness, sentiments, and intellectual labour of the creator. This idea closely aligns with “The Romantic theory of authorship”, which sees the author as a unique, inspired creator- “a genius who produces original work from within”.  The copyright laws in different jurisdictions follow this theory and consider originality as inherently tied to human agency, where the emphasis is on the author’s intellect and not on mechanical labour.

Some scholars argue that application of this theory to the modern creativity process may lead to ignorance of technology – assisted works, including AI- generated works, where human contribution may be indirect or limited or in collaboration with machines. Further, refusing copyright protection to AI-assisted work would place them in the public domain, thereby allowing unrestricted access to such works without the fear of facing copyright infringement claims.   

On the other hand, an argument persists that the determination originality requirement in AI-Generated content would be difficult as the autonomy in AI systems may raise the doubts over the degree of human involvement and application of skill and judgment on part AI user and which has to be decided case by case basis and tool to tool basis.   

Prompt Engineering as a model Threshold of Modern Creativity

Prompt refers to the input given to a generative AI model to obtain a desired output. The quality, clarity, and exclusivity of a prompt significantly affect the reliability of the AI’s output, and this skill of crafting effective prompts to have reliable outputs is known as “prompt engineering” As discussed earlier, under copyright law, originality and human skill are the key requirements to avail protection; in this context, drafting complex prompts, involving multiple iterations, contextual layering, and domain-specific knowledge, reflects a significant degree of intellectual labour and creative decision-making.  This argument is gaining traction, particularly as many academic institutions  and renowned tech corporations like IBM and Microsoft are offering specialised courses and certifications in prompt engineering, treating it as a constructive and distinct skill set.

Further, a similar approach is evident in China through multiple judgments in this context. Since 2019, Chinese courts have decided numerous notable cases concerning the copyrightability of AI- generated content. In the landmark 2019 case “Shenzhen Tencent v. Shanghai Yingxun” , the Nanshan District People’s Court in China addressed the copyrightability of AI-assisted works. The court held that although the article was produced by an AI software, Dreamwriter, it involved substantial human input, such as data selection, algorithm training, and template design, thereby demonstrating intellectual effort and creative expression and is eligible for copyright protection.

In another case in November 2023, “Li v. Liu” , the Beijing Internet Court, while dealing with a case concerning AI – generated images, extended copyright protection to AI – generated content. The court considered efforts on the part of Mr Li that he repeatedly considered what parameters to input into Stable Diffusion (image-generating tool) to have the program generate the final image he wanted.

Another case of 2025, “Wang v Wuhan Technology Company” .  The court considered that Wang enjoyed copyright in the picture he generated via AI, as the picture reflected his aesthetic choices and personal judgment.

Considering the above arguments and the judicial precedents set by the Chinese courts recognising human input in AI-assisted creation, the skill and creative judgment exercised through iterative prompting can satisfy the originality requirement under copyright law.

Who shall be the Owner of AI Generated Work – AI, user or developer?

Historically, copyright laws across most of the jurisdictions have accepted only humans or natural persons as the rightful owners of copyrighted works. “Original works of authorship” , is a term consistently interpreted by Courts to mean a human author (as in the case of “Naruto v. Slater” , where the US court held that a monkey could not claim a copyright over a selfie it had taken, as the law does not recognise non-human entities as owners). In another recent case of 2025, which directly relates to AI-Generated works, “Thaler v. Perlmutter” , the court upheld the US Copyright Office’s rejection of granting ownership to AI, emphasising that human authorship is a fundamental requirement.

In the Indian legal context, in the case of Ankit Sahni v. Registrar of Copyrights (Re: SURYAST), the Copyright Office reassessed the registration of an AI-generated painting after the AI system was listed as a co-author. The case highlights that Indian copyright law currently acknowledges only human authorship, not AI authorship.

Conclusion

It is established that a common thread runs through copyright laws across numerous jurisdictions: creativity, intent, and accountability are human attributes, and copyright law exists primarily to incentivise human authors. Extending authorship to AI would undermine the objective of copyright law, which is to reward human ingenuity and labour. Providing authorship to AI systems could create legal ambiguity, as AI lacks legal personhood, intention, or moral consciousness.

Accordingly, AI-generated works should be regarded as subject to copyright protection only when satisfying the requirement of originality where the creation reflects meaningful human skill, judgment, and creative direction. Such works should therefore remain subject to copyright protection, with the “natural person” who exercises such creative control-the prompter, rather than the AI developer, being recognised as the author Alternatively, if Indian law is hesitant to confer copyright authorship in such circumstances, the “Ukrainian approach” of recognising sui generis rights, particularly economic rights, in AI-generated works may provide a viable alternative for India.

The author is an Assistant Professor (Selection Grade) at UPES School of Law, Dehradun. He can be reached at yatin.kathuria@upes.ac.in