"As artificial intelligence increasingly generates text, music, images and software code, a critical legal question arises: who owns the copyright in AI-generated works? This article examines the position under Indian copyright law, including whether AI can be recognised as an author, the extent of human input required for copyright protection, the originality challenges posed by machine-generated content, and the legal implications of using copyrighted material to train AI systems. The article also explores emerging policy debates in India and outlines practical safeguards for businesses using generative AI."

INTRODUCTION

Artificial intelligence has evolved rapidly from a back-end technological tool into an active and accessible means of creating text, music, images, software code and design.

This transformation raises a foundational legal question: when an AI system generates a literary, musical or artistic work, who owns the copyright? Is it the individual who enters the prompt, the developer of the model, the owner of the datasets used for training, the machine itself, or some combination of these actors?

For businesses deploying generative AI in India, this is no longer a theoretical debate. It directly affects ownership, licensing, enforcement, valuation of intellectual property assets and regulatory risk.

To examine these issues, it is useful to begin with the traditional legal understanding of authorship.

WHO IS AN “AUTHOR”?

India is a signatory to the Berne Convention for the Protection of Literary and Artistic Works, which remains the cornerstone of international copyright protection. Although the Convention does not expressly define an “author” as a human being, its structure strongly assumes personhood.

For instance, Article 3 refers to authors by nationality and habitual residence, while Article 7 links copyright duration to the author’s life and a period after death.

These concepts naturally contemplate a human or legally recognised person as the author of copyrightable works.

This becomes significant in the AI context. If an AI system were itself treated as the author, traditional copyright doctrines relating to ownership, term, succession, moral rights and enforcement would become difficult to apply in their present form.

THE INDIAN CONTEXT

In India, the starting point is the Copyright Act, 1957. Section 2(d) provides that, in relation to a computer-generated literary, dramatic, musical or artistic work, the “author” is the person who causes the work to be created.

This is the key provision governing AI-generated content. It recognises that technology may be used to create copyrightable works, but attributes authorship to a “person”.

CAN AI BE AN “AUTHOR”?

As the law presently stands, this is unlikely. An AI system has no independent legal personality under Indian law. It cannot own property, assign rights, sue or be sued, or exercise moral rights.

IF NOT THE AI, THEN WHO OWNS THE COPYRIGHT?

Authorship is likely to vest in the person or entity that can show they caused the work to be created. Depending on the facts, this may include:

• the individual user directing the output;
• an employer or commissioning party; or
• a company acting through its employees or agents.

The question is not who owns the software, but who exercised meaningful creative control over the final work.

HOW IS THIS DETERMINED?

Indian courts are likely to consider the level of human involvement, including:

• who conceived the idea or intended result;
• whether prompts were detailed and iterative;
• whether outputs were selected or curated;
• whether the result was edited or transformed; and
• who controlled the overall creation process.

The stronger the evidence of human judgment, skill and control, the stronger the claim to authorship.

WHAT IS THE STANDARD FOR COPYRIGHT PROTECTION?

Even where authorship is attributed to a person, the work must still satisfy the requirement of originality under Indian law. In practical terms, the claimant should be able to show:

• Human intellectual contribution beyond a purely automated output;
• Original expression involving skill, judgment or creative choices; and
• A clear causal connection between the claimant and the creation of the work.

If the output is entirely machine-generated with negligible human input, protection may be harder to establish.

PRACTICAL POSITION IN INDIA

Indian law currently does not ask whether the machine created the work. It asks which person caused the work to be created, and whether that person contributed sufficient originality and control to justify copyright protection.

AI may generate the content, but authorship is still likely to rest with the human or entity directing the creative process.

USE OF COPYRIGHTED WORKS FOR AI TRAINING: THE EMERGING INDIAN DEBATE

Vast data sets contained copyrighted works are used to train the AI models. What is the legality of using such datasets?

Section 14 of the Copyright Act, 1957 grants rightsholders exclusive rights to reproduce works and create adaptations. Depending on how AI training occurs, the use of copyrighted content may infringe these rights. How can this be remedied? The current policy debate broadly reflects two competing positions.

Many technology companies seek an exception under copyright law permitting broader use of copyrighted material for training datasets. Such an approach may accelerate AI development, but it may also weaken core copyright protections.

Artists, authors and creative industry associations have advocated for a licensing regime under which technology companies must obtain permission and compensate rightsholders for the use of copyrighted works. While protective of creators, a purely transaction-based system may be commercially burdensome, slow and difficult to scale.

A Possible Middle Path: The Hybrid Model

As referenced in a working paper reportedly issued by the Department for Promotion of Industry and Internal Trade on 8 December 2025, a possible solution may lie in a Hybrid Model.

Under such a framework, copyright holders may not be able to block training use entirely; however, developers may be required to pay royalties through a centralised collection mechanism, akin to a copyright society constituted for this purpose.

Whether such a model proves workable in practice will depend on pricing, transparency, enforcement and administrative efficiency.

RECOMMENDED SAFEGUARDS FOR BUSINESSES

For businesses using generative AI in India, the immediate legal priority should be risk management.

• Contractual Ownership Clauses - Ensure employee, consultant and vendor agreements clearly assign rights in AI-assisted outputs.
• Prompt Governance - Maintain records of prompts, revisions and human editorial involvement.
• Human Review Layers - Substantial human curation may strengthen authorship claims.
• Dataset Compliance - Use licensed, open-source or lawfully obtained datasets wherever possible.
• Similarity Screening - Review outputs for infringement risks before commercialisation.

CONCLUSION

Indian law currently treats AI as a sophisticated tool rather than a legal creator. While machines may generate commercially valuable outputs, authorship under the present framework remains tied to human responsibility, human creativity and legal accountability.

Unless future legislation expressly recognises autonomous machine authorship or grants AI some form of legal personality, copyright in India is likely to remain vested in the individuals or entities that cause such works to be created.