From Prompt to Protection: Evaluating Copyright Ownership of AI-Generated Works in India

Written by: Shradha S. Prasad, B.A.LL.B (Hons.) | Christ University Pune Lavasa 

JUSTICE

 Introduction

Copyright promotes innovation by giving authors exclusive rights over their works. Authorship normally is human creativity. But now, with the arrival of generative AI models such as DALL-E, Midjourney, Stable Diffusion or even Indian developed AI models, we can prompt and generate literary, artistic and musical works, prompting questions. Is a work created mostly by an AI "original"? And whose work is it? the prompter, the AI developer, the AI, or just anyone? And, who owns the copyright?

These questions are critical in the burgeoning AI and digital economy in India. Generative AI is not mentioned in the Copyright Act, 1957. [1]Section 13 protects original literary, dramatic, musical and artistic works, but requires originality and authorship. [2]The author of a literary, dramatic, musical or artistic work generated by a computer is defined as "the person who causes the work to be created" under Section 2(d)(vi) of the Act, which was added in 1994[3]. This was drafted in terms of computer rule-based programs, rather than neural networks that generate works. This paper analyses the authorship of works created by AI in India through statutory, case, doctrinal and comparative law approaches. It concludes that AI-generated works, without significant human creative input, are unlikely to be protected by copyright law, but works with significant human input may be protected with the human as the copyright owner. Incentives are needed to encourage innovation, and to compensate human authors whose works are used for training AI.

Indian Copyright Act, 1957

Indian copyright is human-centred. "Author" is defined in Section 2(d) as:

- In the case of a literary, dramatic or musical work: person writing or composing;

- In the case of other artistic works (not photographs): the artist;

- In the case of a computer-generated work: the author.

Under Section 17, author is the first owner of copyright (subject to a few exceptions, such as work of an employee). [4]In India, a person would be a natural person or juristic person (a company), but not a machine or AI, which are not persons. Authorship has been seen as a human endeavour. Section 13 demands "originality". India has created the concept of originality. The previous "sweat of the brow" (mere labour) test was modified in Eastern Book Company v. D.B. Modak (2008) [5]where the Supreme Court applied a "modicum of creativity" or "skill and judgement" test. Originality requires the work is the author's own, mental product and not a copy, but not new. The human test doesn't apply to AI works, which statistically combine features of their training data.[6]

As for infringement, Section 51 has unauthorised copying or adaptation. [7]Section 52 provides for fair dealing for research, criticism or review. [8]Training AI with copyright works (text and data mining) is a grey area in Indian jurisprudence as to whether it is fair dealing or infringement but there might be arguments on transformative use based on cases such as R.G. Anand v. Deluxe Films (1978) [9]where the idea was separated from expression. The computer-generated work provision in the Act may help AI works. The human prompter or developer could be an author if they "caused" the work to be created (by the prompts, selections and editing). But it may not help autonomous AI works with little human input, they may be either unprotected or in the public domain.[10]

 Key Judicial Precedents on Originality and Authorship

Creativity is a human specialty. In the case of Deluxe Films v R.G. Anand (AIR 1978 SC 1613), The Supreme Court agreed with the idea expression dichotomy.[11] It's infringement, but not creative infringement. The idea-expression dichotomy will not apply to AI; a different composition will remain creative (not based on an AI composition created by humans). Neither pure labour (human decisions) nor other (non-creative) decisions (such as style, layout, and editing) are unique to humans. But editing, story and ideas will be standards. Other circumstances show human author. Existing precedents on training AI such as Syndicate of the Press of the University of Cambridge v. B.D. Bhandari [12]and transformative uses mean that non-expressivity uses (transformative uses) will not be infringing and will be weighed in. But in 2026, the Supreme Court doesn't have any cases regarding generative AI authorship. The new cases are decided on facts and merits by the Copyright Office and district courts.

 India's Test Case - RAGHAV

The most prominent case in India is the RAGHAV case (artist-lawyer or author "Ankit Sahni" and AI tool RAGHAV - [13]Responsive Artificially Generated High Art Visualizer or Robust Artificially Intelligent Graphics and Art Visualizer). Sahni has applied for copyright in 2020 (he used his picture and RAGHAV to create an art work similar to Van Gogh’s “Starry Night" with AI tool). The Indian copyright office did not grant copyright. In November 2020, Sahni secured a copyright (co-authorship of Sahni and RAGHAV) and it is the world's first co-authorship. But recently, this work has been withdrawn by the Office since it has "questioned the position of RAGHAV as 'author' under Section 2(d)(iii) and (vi) and the human effort put into creating the work".

Is the work (style and imagery prompts) creative? This work is not registered in US; they claimed that there is no actual human effort. The same is registered in Canada (Sahni as author). In India, the copyright office initially granted a registration listing both Sahni and Raghav as co- authors, marking the world first such co- authorship. This discrepancy suggests that in the eyes of many regulators, simply “teaching” or prompting an AI is not sufficient to claim copyright, there must be a level of human creativity involved.  The Indian Copyright Offices reversal reflects a shift toward the view that human must be a primary creative force, rather than a computer tool.

Parallel to the RAGHAV Case, Stephen Thaler has sought copyright in India for a Recent Entrance to Paradise which is a work create by his AI system, DABUS. [14]Following the directive from Delhi High Court, the copyright office fast tracked this application, which notably lacked a direct human creativity. There is a growing concern among legal experts that if Indian authorities are too permissive in granting AI copyrights, it could undermine the rights of human artist whose work are often use to train these models, leading to potential infringement issues.

Comparative Perspectives

UK Copyright, Designs and Patents Act 1988 (s.9(3))[15] expressly provide the author of copyright works created using computer is the “person by whom the arrangements necessary for the creation of the work are undertaken”.  US Copyright is granted only if human is involved in the creative process.[16] Works are copyrightable if a human is on the hook through significant selection or editing, but they remain uncopyrightable if there is no human effort.  In China and EU, they require human to be on the hook to meet the standard of originality.  In India the legal position remains unclear and potentially inconsistent. While the department for the promotion of industry and internal trade released working paper on generative AI and Copyright in 2025, it focused on hybrid sites of training and licensing data rather than establishing rules for the ownership of AI output or the creation of new IP.

Challenges and Policy Considerations

The current debate surrounding AI- generated content presents significant challenges in intellectual property policy, concerning whether the lack of protection for AI- only works might lead to a decline in the creative output or if the protection should be applied during the training phase. Legal experts suggests that ownership models will likely involve a hybrid of data sources, user prompts, and humans’ authors, though proving infringement remains very difficult. [17]Ethically while non-commercial research training may fall under fair dealing via section 52, commercial training for Large Language Models necessitates specific safe harbour and licensing frameworks[18]. This tension is currently being tested in the Delhi High Court through the ANI v OpenAI case, [19]which centres on the unauthorized use of news articles for model training.

Recommendations

The primary recommendation focuses mainly on the legislative and regulatory clarity, suggesting that the Copyright Act be amended. Or the guidelines issued by the Copyright Office to mandate that any protected work demonstrate significant human creativity, such as through specific prompts, refinements, or editing. [20]Under this framework, AI will be barred from being recognized as author or co-author, reinforcing that” human who causes “principle where authorship is granted to the individual directing the creative process rather than a mere button pusher. In order to support this proper registration process is to be followed where applicants have to disclose both AI and human inputs, with the authority to deny applications that lack sufficient human contribution. [21]

Furthermore, the framework should address all the complexities of training data by putting out clear rules, whether through text or data mining exceptions, opt out mechanisms of collective licensing protections. It is also recommended that India align with the global standards while also tuning policies as per its domestic needs. Ultimately a hybrid solution is suggested to meet the interest of the Public.

Conclusion

The state of Indian Copyright law is considered ill equipped to handle the complexities og Generative AI, spanning the entire process from initial prompting to final protection. The case of RAGHAV very clearly demonstrates that the existing provisions for computer generated works are inappropriate for systems that learn autonomously rather than merely following static instructions. While section 2(d) is broad, Indian legal cases such as Eastern Book Company and R.G.Anand throws light on the fact that works must be possess human creativity to be eligible for protection. Consequently, there is likely no copyright and therefore no enforceable rights in purely computer-generated works, although potential human rights may exist for hybrid works.

Failure to update the legal framework may result is misuse of technology. It is herby argued that India must clarify ownership rules in order to prevent AI from being erroneously categorized as a non-human author. Current practices reflect that the law requires a specific tweaking to remain relevant. Ultimately a balance must be struck that rewards creativity while remaining consistent with evolving technology to support growth of Indias AI and creative industries.

[1] K. Singhania, Regulating AI in India: Challenges, Initiatives, and Path to Future Success, Mondaq (May 7, 2025), https://www.mondaq.com/india/new-technology/1621322/regulating-ai-in-india-challenges-initiatives-and-path-to-future-success. (guides.library.harvard.edu)

[2] The Copyright Act, 1957, § 13, No. 14, Acts of Parliament, 1957 (India).

[3] The Copyright Act, 1957, § 2(d)(vi), No. 14, Acts of Parliament, 1957 (India) (as amended in 1994).

[4] The Copyright Act, 1957, § 17, No. 14, Acts of Parliament, 1957 (India).

[5] E. Book Co. v. D.B. Modak, (2008) 1 S.C.C. 1 (India).

[6] Recovering Personality in Copyright’s Originality Inquiry, 138 Harv. L. Rev. (2025), https://harvardlawreview.org/print/vol

[7] The Copyright Act, 1957, § 51, No. 14, Acts of Parliament, 1957 (India).

[8] The Copyright Act, 1957, § 52, No. 14, Acts of Parliament, 1957 (India).

[9] R.G. Anand v. Deluxe Films, (1978) 4 S.C.C. 118 (India).

[10] Adnan Masood, Intellectual Property Rights and AI-Generated Content — Issues in Human Authorship, Fair Use Doctrine, and Output Liability, Medium (Apr. 4, 2025), https://medium.com/

[11] R.G. Anand v. Deluxe Films, (1978) 4 S.C.C. 118 (India).

[12] Syndicate of the Press of the Univ. of Cambridge v. B.D. Bhandari, AIR 2011 Del. 98 (India).

[13] Altacit Global, AI-Generated Works in India: Examining the “Raghav” Signal in Copyright Law, Medium (Jan. 31, 2026), https://medium.com/

[14] Khadija Khan, AI-generated art covered by India’s copyright law? Delhi HC gives copyright office 8 weeks to decide, ThePrint (Apr. 10, 2026, 6:30 AM IST), https://theprint.in/

[15] Copyright, Designs and Patents Act 1988, c. 48, § 9(3) (UK).

[16] Copyright Act of 1976, 17 U.S.C. §§ 101–1511 (2024).

[17] Lawful Legal, The New Hybrid Model for AI & Copyright (2025): Finding the Middle Ground, Lawful Legal (Dec. 24, 2025), https://lawfullegal.in/the-new-hybrid-model-for-ai-copyright-2025-finding-the-middle-ground/

[18] Khadija Khan, AI-generated art covered by India’s copyright law? Delhi HC gives copyright office 8 weeks to decide, ThePrint (Apr. 9, 2026), https://theprint.in/judiciary/ai-generated-art-covered-by-indias-copyright-law-delhi-hc-gives-copyright-office-8-weeks-to-decide/2901029/

[19] SpicyIP, [Part I] ANI v. Open AI – A Lesson in Resisting the Temptation to Borrow Excessively without Legislative Sanction, SpicyIP (May 30, 2025), https://spicyip.com/2025/05/part-i-ani-v-open-ai-a-lesson-in-resisting-the-temptation-to-borrow-excessively-without-legislative-sanction.html

[20] The Requirements for Copyright Protection, HARVARD CYBERLAW SCH., https://cyber.harvard.edu/cx/The_Requirements_for_Copyright_Protection (last visited May 1, 2026).

[21] Revised AI Advisory: How AI Regulation Can Impact India, INDIAAI, https://indiaai.gov.in/article/revised-ai-advisory-how-ai-regulation-can-impact-india (last visited May 1, 2026).