RESEARCH PAPER: ARTIFICIAL INTELLIGENCE AND COPYRIGHT LAW IN INDIA, AUTHOR: PRIYA SANDHU
The rapid development of generative artificial intelligence (AI) has unsettled conventional assumptions of copyright law, particularly the idea that creative expression is attributable to a human author. AI systems can now generate literary text, music, images and audiovisual material, while their training may involve the collection and processing of copyright-protected works. This paper examines whether India’s Copyright Act, 1957 is equipped to address these developments. It analyses the statutory meaning of authorship, the provision governing computer-generated works, the originality standard developed by the Supreme Court, and the statutory exceptions relevant to AI training. It also considers the Delhi High Court’s interim decision in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC and the Government of India’s 2025 consultation initiative, while comparing selected approaches in the United Kingdom and the United States. The paper adopts a doctrinal and comparative methodology based on legislation, judicial decisions, official policy material and secondary sources
RESEARCH PAPER
ABSTRACT
The rapid development of generative artificial intelligence (AI) has unsettled conventional assumptions of copyright law, particularly the idea that creative expression is attributable to a human author. AI systems can now generate literary text, music, images and audiovisual material, while their training may involve the collection and processing of copyright-protected works. This paper examines whether India’s Copyright Act, 1957 is equipped to address these developments. It analyses the statutory meaning of authorship, the provision governing computer-generated works, the originality standard developed by the Supreme Court, and the statutory exceptions relevant to AI training. It also considers the Delhi High Court’s interim decision in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC and the Government of India’s 2025 consultation initiative, while comparing selected approaches in the United Kingdom and the United States. The paper adopts a doctrinal and comparative methodology based on legislation, judicial decisions, official policy material and secondary sources. It argues that Indian law can accommodate some AI-assisted works where meaningful human creative contribution is identifiable, but it does not yet provide a sufficiently clear framework for autonomous outputs or large-scale training practices. The paper recommends transparent disclosure of AI involvement, clearer standards for human contribution, proportionate licensing mechanisms and safeguards for both creators and technological development.
Keywords:
Artificial Intelligence; Copyright; Authorship; Originality; Generative AI; Copyright Act, 1957.
1. INTRODUCTION
Copyright law has traditionally been built around human creativity. Authors write books, composers create music and artists produce paintings. Generative artificial intelligence has complicated this familiar relationship. By processing large datasets and identifying patterns, modern AI systems can produce text, images, music, software code and other expressive material in response to user instructions. The resulting work may appear creative even where the human user has contributed only a short prompt.
This development raises two connected but distinct questions. First, can material generated with or by AI qualify for copyright protection, and if so, who should be recognised as its author and owner? Secondly, may developers use copyright-protected works to train AI systems without obtaining permission from rightsholders? The first question concerns the output of AI systems; the second concerns the inputs used to develop them. Treating these issues as identical would obscure the different rights and legal tests involved.
In India, the principal legislation is the Copyright Act, 1957. The Act was not drafted with contemporary generative models in mind, but it contains a provision concerning computer-generated works. Its interpretation, particularly alongside the statutory requirements of originality and authorship, is now central to the debate. The recent litigation in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC and the Government’s consultation on generative AI and copyright demonstrate that the issue is no longer merely theoretical.1
This paper argues that India should preserve the central role of human creative contribution while clarifying how the law applies to AI-assisted works, autonomous outputs and training datasets. A balanced framework should protect genuine authors without unnecessarily obstructing research, innovation and access to knowledge.
2. RESEARCH PROBLEM AND QUESTIONS
The central research problem is the uncertainty created by applying a human-centred copyright statute to systems capable of generating expressive material with varying degrees of human involvement. Section 2(d)(vi) identifies the author of a computer-generated work as the person who causes it to be created, but the Act does not explain how this test should operate when a user supplies prompts, a developer designs the model and the system independently generates the final expression.2
The paper addresses four questions: (i) whether AI-generated and AI-assisted works satisfy the requirements of copyright protection in India; (ii) who should qualify as the author and first owner of such works; (iii) whether the use of protected works for AI training may fall within existing exceptions or requires licensing; and (iv) what legal and policy measures could provide certainty to creators, users and AI developers.
3. RESEARCH OBJECTIVES AND METHODOLOGY
The objectives are to examine the relevant provisions of the Copyright Act, analyse Indian judicial principles on originality and infringement, assess emerging disputes involving AI training, compare selected foreign approaches and propose practical reforms for India.
The study adopts a doctrinal legal research methodology. It relies on primary sources, including the Copyright Act, 1957, judicial decisions and official policy documents, together with secondary legal commentary and comparative material. The analysis is qualitative and interpretative; it does not claim to measure the economic effects of AI training or to establish empirical facts about the volume of copyrighted material used by particular systems.
4. THE INDIAN COPYRIGHT FRAMEWORK
4.1 Copyrightable subject matter and originality
Section 13 of the Copyright Act provides for copyright in original literary, dramatic, musical and artistic works, as well as cinematograph films and sound recordings. Section 14 defines the exclusive rights associated with copyright, including reproduction and communication to the public. The statutory scheme therefore protects qualifying expression rather than abstract ideas, facts or styles.3
The Supreme Court’s decision in Eastern Book Company v. D.B. Modak is particularly relevant to AI-generated material. The Court rejected a purely labour-based approach and required a minimum degree of creativity, involving skill and judgment, for a work to be original. The decision does not directly address AI, but it provides a useful test: the investment of time, money or effort alone is not enough; the work must contain a sufficient element of original intellectual contribution.4
In R.G. Anand v. Delux Films, the Supreme Court explained that copyright does not protect an idea, theme or subject as such; it protects the particular form of expression. This distinction is important where an AI output resembles a protected work. Similarity in a general concept or style is not automatically infringement, but reproduction of a substantial part of protected expression may be actionable.5
4.2 The statutory author of a computer-generated work
Section 2(d)(vi) of the Act states that, in relation to a literary, dramatic, musical or artistic work which is computer-generated, the author is the person who causes the work to be created. The provision is significant because it anticipates that a computer may be involved in producing a work. However, it does not define “computer-generated” or specify whether the relevant person is the user, programmer, model developer, operator or another person who arranged the process.
A distinction should be drawn between an AI-assisted work and a substantially autonomous AI output. In an AI-assisted work, a human may make original choices about the subject, composition, structure, selection, editing and final arrangement. The AI functions as a tool, much like other digital technologies. Where those choices are sufficiently creative and identifiable in the final work, copyright may subsist in the human contribution. By contrast, where a user merely enters a broad instruction and accepts an output generated autonomously, it may be difficult to identify the human expression required by the originality standard.
The phrase “person who causes the work to be created” should not be treated as automatically granting copyright to every person who initiates an AI process. Causation is broader than authorship in the creative sense. A developer may have built the model, a company may provide the service and a user may type the prompt, but none of those facts alone necessarily establishes that the individual contributed original expression to the particular output. The statutory provision needs interpretation that considers control, creative choices and the relationship between the human contribution and the final work.
5. AUTHORSHIP AND OWNERSHIP OF AI-GENERATED WORKS
5.1 Can an AI system be an author?
Under the present Indian legal framework, an AI system cannot independently hold copyright as an author. The Act uses the concept of a “person” and its ownership provisions contemplate legally recognised persons capable of holding and transferring rights. An AI model is a technological system, not a natural or juristic person. Recognising it as an author would also raise unresolved questions about legal capacity, duration of rights, enforcement, liability and the exercise of moral rights.
A reported decision of the Registrar of Copyrights dated 31 August 2026 in the application concerning “A Recent Entrance to Paradise”, attributed to the AI system DABUS, reinforces this distinction. The Copyright Office reportedly accepted that the work could meet the threshold of originality on the material before it, but rejected DABUS as the named author because it was not a natural or juristic person recognised by law. The decision is an administrative registration determination, not a binding judgment of a constitutional court; nevertheless, it is a significant indication of the current administrative approach.6
5.2 Identifying the human contribution
The more defensible approach is to assess the human contribution on a case-by-case basis. Relevant factors may include the specificity and creative character of instructions, iterative prompting, the ability to reject and refine outputs, substantive editing, selection among generated alternatives and the creative arrangement of material. These factors should be evidence of actual creative control, not a checklist that automatically confers ownership.
A user who substantially rewrites an AI-generated draft, selects original elements and creates a distinctive arrangement may claim copyright in those human-authored elements. Such protection should not extend automatically to portions generated without sufficient human authorship. Similarly, a developer should not become the author of every output merely because the developer designed the model or owns the underlying software. Contractual terms may allocate rights between users and service providers, but a contract cannot create copyright in material that does not qualify for protection under the statute.
5.3 First ownership, assignments and moral rights
Section 17 generally identifies the author as the first owner of copyright, subject to statutory exceptions, including certain employment and commissioned-work situations. Sections 18 and 19 govern assignment and its requirements. These provisions assume that a legally identifiable author or first owner exists. In AI-assisted projects, parties should therefore specify who is responsible for human-authored contributions, what rights are being assigned and whether third-party material has been incorporated.7
Moral rights under Section 57, including the right to claim authorship and to restrain certain prejudicial distortions, are also difficult to apply to autonomous outputs. An AI system cannot meaningfully exercise personal attribution or integrity interests. The law should avoid attributing moral rights to a machine while ensuring that identifiable human contributors are not deprived of recognition.
6. COPYRIGHT INFRINGEMENT AND AI TRAINING DATA
6.1 The input-side problem
Generative AI models are trained on large collections of data. Such datasets may include books, news articles, photographs, music, illustrations and other material protected by copyright. Training may involve downloading, copying, storing, preprocessing and analysing works. Since Section 14 includes reproduction and electronic storage within the exclusive rights of copyright owners, the use of protected material during training can raise infringement questions, depending on the acts performed, the rights involved and the availability of a statutory exception.8
The fact that a work is publicly accessible on the internet does not, by itself, mean that it is free of copyright or available for unrestricted commercial use. At the same time, not every technical interaction with a work necessarily results in actionable infringement. The legal analysis must identify the relevant restricted act and consider whether an exception applies.
6.2 Fair dealing under Section 52
Section 52(1)(a) recognises fair dealing for specified purposes, including private or personal use, research, criticism or review, and reporting current events. India’s provision is a purpose-specific fair-dealing framework, not an unrestricted general fair-use clause. Whether AI model training falls within it requires careful analysis of the purpose and fairness of the particular use. Commercial scale, the nature of the works, the amount taken, market effects and the technical functioning of the system may all be relevant to the assessment.9
The question is not resolved simply by describing AI training as research or innovation. A commercial system may have research and public-interest benefits, but those benefits must be considered alongside the statutory rights of authors and publishers. Equally, a categorical rule that every use of protected material in machine learning is infringement could impede legitimate research and technological development. The law requires a principled balance.
7. RECENT DEVELOPMENTS: ANI v. OPENAI AND GOVERNMENT CONSULTATION
7.1 ANI Media Pvt. Ltd. v. OpenAI OpCo LLC
The dispute between ANI Media Pvt. Ltd. and OpenAI is a major Indian proceeding concerning the use of copyrighted material to train a generative AI model. ANI alleged, among other things, that its news content had been used without authorisation and sought interim relief. On 24 July 2026, a single judge of the Delhi High Court declined to grant the requested interim injunction. At the interim stage, the Court considered the purpose and fairness of the use and expressed a prima facie view that the relevant training use could fall within Section 52(1)(a). It also considered whether the outputs substantially reproduced ANI’s protected expression.10
The decision must be read with care. It concerned an application for interim relief, not a final determination after trial of every issue in the suit. The Court expressly indicated that its observations would not determine the final outcome. ANI challenged the order, and the Division Bench issued notice in September 2026; the appeal remained pending as of 30 October 2026, with a further hearing scheduled for December 2026.11
The case illustrates the difficulty of applying a purpose-specific fair-dealing provision to large-scale AI training. It also raises questions about jurisdiction, the location of training activities, the role of retrieval systems and the relationship between model training and the reproduction of protected content in outputs. The final appellate and trial outcomes will be important in shaping Indian law.
7.2 Government working paper on generative AI and copyright
On 8 December 2025, the Department for Promotion of Industry and Internal Trade notified the formation of a committee to examine whether India’s existing copyright framework adequately addresses generative AI or requires amendment. The notification circulated the committee’s Working Paper—Part I, focusing on the use of copyright-protected works for AI training, for stakeholder consultation. This demonstrates that the Government recognises the need to examine the issue systematically rather than rely solely on litigation.12
A consultation process is not itself a change in law. Any proposed licensing or remuneration framework must be assessed against the interests of creators, publishers, technology companies, researchers and the public. The policy challenge is to provide lawful access to training material without making compliance so burdensome that smaller developers and research institutions are excluded.
8. COMPARATIVE PERSPECTIVES
8.1 United Kingdom
Section 9(3) of the United Kingdom’s Copyright, Designs and Patents Act 1988 provides that, for a computer-generated literary, dramatic, musical or artistic work, the author is taken to be the person by whom the arrangements necessary for creation are undertaken. This provision offers a statutory attribution rule, although questions remain about how it applies to modern generative systems and who made the relevant arrangements. India’s Section 2(d)(vi) is similar in referring to the person who causes the work to be created, but both provisions require careful application to complex AI workflows.13
8.2 United States
The United States Copyright Office’s January 2025 report on copyrightability concluded that copyright protects human-authored expression, including human-authored elements within works that also contain AI-generated material. It stated that purely AI-generated material is not protected where there is insufficient human control over expressive elements, and that prompts alone will not ordinarily provide sufficient control under currently available technology. The report favours case-by-case assessment of human contribution rather than a blanket exclusion of all works created with AI.14
The American approach is not binding in India, but it offers a useful analytical model. It separates the use of AI as an assistive tool from the replacement of human creative choices. India could adopt a similarly contribution-focused test while interpreting Section 2(d)(vi) in light of its own statutory language and jurisprudence.
9. FINDINGS
The research produces five principal findings. First, the Copyright Act, 1957 does not expressly regulate generative AI, but Section 2(d)(vi) provides a starting point for identifying authors of computer-generated works. Its application to modern systems remains uncertain. Second, the originality standard in Eastern Book Company v. D.B. Modak makes it difficult to justify protection based solely on labour, investment or a minimal prompt; identifiable creative contribution remains important. Third, AI-assisted works may qualify for protection to the extent that original human expression can be identified, whereas fully autonomous outputs present a more difficult claim. Fourth, AI training may involve acts covered by copyright owners’ exclusive rights, but the applicability of Section 52 depends on a fact-specific fair-dealing analysis. Fifth, recent developments—including ANI v. OpenAI and the DPIIT consultation—show that Indian law is actively confronting these questions, but important issues remain unsettled.
10. SUGGESTIONS AND RECOMMENDATIONS
First, the legislature or the Copyright Office should issue clear guidance distinguishing AI-assisted works from substantially autonomous outputs. The guidance should explain how human creative contribution may be assessed, without treating prompt length or technical complexity as a substitute for originality.
Secondly, applicants seeking registration should be encouraged to disclose material AI involvement and identify the human-authored elements claimed for protection. Such disclosure would improve transparency and help prevent overbroad claims, while avoiding unnecessary disclosure of confidential business information.
Thirdly, India should consider a transparent and workable licensing mechanism for the use of protected works in commercial AI training. Any mechanism should include meaningful participation by creators and collective management organisations, reasonable remuneration where appropriate, clear record-keeping and practical safeguards for small developers and non-commercial research.
Fourthly, the law should clarify the treatment of text-and-data-mining activities and the relationship between training, storage and output-stage reproduction. A narrowly designed exception or licensing framework should specify its scope, safeguards and interaction with existing Section 52 exceptions. It should not be assumed that all training is automatically fair dealing or automatically infringement.
Finally, contracts between AI providers and users should clearly address ownership of human contributions, permitted uses, warranties concerning third-party material and responsibility for infringing outputs. Contractual clarity cannot replace statutory reform, but it can reduce disputes while the law develops.
11. CONCLUSION
Generative AI challenges copyright law because it separates the production of expressive material from the familiar process of human creation. India’s Copyright Act, 1957 contains provisions that can address parts of this challenge, particularly the definition of a computer-generated work and the existing rules on originality, ownership and fair dealing. However, those provisions were not designed for systems trained on enormous datasets and capable of generating content with limited direct human intervention.
The better approach is neither to grant copyright automatically to every AI output nor to deny protection to every work involving AI. Copyright should remain available where a human makes a meaningful and identifiable creative contribution, while purely autonomous output should not receive protection merely because it is commercially valuable or technically sophisticated. The use of protected works for training must be examined separately under the exclusive rights and exceptions in the Act. ANI v. OpenAI may influence that debate, but its interim findings and pending appeal should not be mistaken for a final statement of Indian law.
India now has an opportunity to develop a balanced framework that rewards human creativity, respects copyright owners, supports responsible innovation and provides certainty to technology developers. Clear guidance, transparent practices and carefully designed licensing or statutory mechanisms can help ensure that technological progress and copyright protection operate in a mutually supportive manner.
12. REFERENCES / BIBLIOGRAPHY
LEGISLATION
Copyright Act, 1957, No. 14 of 1957 (India).
Copyright Rules, 2013 (India).
Copyright, Designs and Patents Act 1988, c. 48 (UK).
Agreement on Trade-Related Aspects of Intellectual Property Rights, Apr. 15, 1994, 1869 U.N.T.S. 299.
CASES
Eastern Book Co. v. D.B. Modak, (2008) 1 SCC 1 (India).
R.G. Anand v. Delux Films, (1978) 4 SCC 118 (India).
ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, CS(COMM) 1028/2024, I.A. 45300/2024 (Delhi High Court July 24, 2026).
OFFICIAL AND POLICY MATERIAL
Department for Promotion of Industry and Internal Trade, Ministry of Commerce and Industry, Government of India, Notification No. P-24029/34/2025-IPR-VII (Dec. 8, 2025).
U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (Jan. 2025).
Registrar of Copyrights, decision concerning Dr. Stephen L. Thaler’s application for “A Recent Entrance to Paradise” (Aug. 31, 2026), as reported in SCC Times (Sept. 1, 2026).
SECONDARY SOURCES
World Intellectual Property Organization, WIPO Lex, Eastern Book Company & Ors. v. D.B. Modak & Anr. (India, Dec. 12, 2007).
World Intellectual Property Organization, WIPO Lex, R.G. Anand v. M/s. Delux Films & Ors. (India, Aug. 18, 1978).
Copyright Act, 1957, No. 14 of 1957, §§ 2(d)(vi), 13, 14, 17, 18, 19, 51–52 (India).
Id. § 2(d)(vi).
Eastern Book Co. v. D.B. Modak, (2008) 1 SCC 1 (India).
R.G. Anand v. Delux Films, (1978) 4 SCC 118 (India).
Copyright Act, 1957, § 52(1)(a)(i) (India).
Department for Promotion of Industry and Internal Trade, Ministry of Commerce and Industry, Government of India, Notification No. P-24029/34/2025-IPR-VII (Dec. 8, 2025) (circulating Working Paper—Part I on Generative AI and Copyright for stakeholder consultation).
ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, CS(COMM) 1028/2024, I.A. 45300/2024 (Delhi High Court July 24, 2026) (interim injunction application; prima facie findings only).
Id. The order was corrected on July 27, 2026; the appeal was pending before a Division Bench as of October 30, 2026.
Copyright, Designs and Patents Act 1988, c. 48, § 9(3) (UK).
U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (Jan. 2025).
Agreement on Trade-Related Aspects of Intellectual Property Rights arts. 9(2), 13, Apr. 15, 1994, Marrakesh Agreement Establishing the World Trade Organization, Annex 1C, 1869 U.N.T.S. 299.
Copyright Act, 1957, §§ 17, 18–19 (India).
Id. §§ 51–52.
Copyright Act, 1957, § 45; Copyright Rules, 2013, r. 70 (India); see also Registrar of Copyrights, decision concerning Dr. Stephen L. Thaler’s application for “A Recent Entrance to Paradise” (Aug. 31, 2026), as reported in SCC Times (Sept. 1, 2026).