Artificial Intelligence and Copyright Protection in India: Challenges under the Copyright Act, 1957

By Ayaan Kakkar

Abstract

The proliferation of generative artificial intelligence systems capable of producing literary, artistic, musical and audiovisual outputs has unsettled the foundational assumptions of Indian copyright law. The Copyright Act, 1957, conceived in an era of exclusively human creation, presumes a natural person as the author of every protected work, with the limited statutory concession in Section 2(d)(vi) for computer-generated works pointing only to the person who causes the work to be created. This paper interrogates the adequacy of that framework against machine-generated expression. It traces the conceptual evolution of authorship, examines the Indian statutory scheme and judicial doctrine on originality, and contrasts the position in India with developments in the United Kingdom, the United States, the European Union, Singapore and Australia. The paper also engages with the Parliamentary Standing Committee on Commerce, Report No. 161 of 2021, the controversial registration of the artwork SURYAST involving the RAGHAV application, and the unresolved questions surrounding training-data ingestion, infringement liability and moral rights. The argument advanced is that India does not yet require a wholly new statute; rather, calibrated amendments to the 1957 Act, supplemented by sui generis protection for fully autonomous output and a narrowly tailored text and data mining exception, would best preserve the constitutional balance between creators, users and the public domain.

Introduction

Copyright law in India has long rested on the premise that creativity is a distinctly human endeavour. The Copyright Act, 1957, which gave statutory expression to the rights of authors in independent India, was modelled substantially on the United Kingdom Copyright Act, 1956, and inherits the same anthropocentric vocabulary. Generative artificial intelligence has displaced that comfortable assumption. Systems such as GPT-4, Stable Diffusion, Mid journey and DALL-E now produce textual, pictorial and musical works whose surface features are often indistinguishable from those crafted by trained human authors. Two distinct legal anxieties follow. The first concerns the eligibility of machine-generated output for copyright protection and the identity of the author. The second concerns the lawfulness of the vast ingestion of pre-existing copyrighted works that is required to train such systems.

The Indian response so far has been hesitant. The Copyright Office registered the painting SURYAST in November 2020 with the RAGHAV Artificial Intelligence Painting Application listed as a co-author alongside Ankit Sahni, only to issue a withdrawal notice in November 2021. The Parliamentary Standing Committee on Commerce, in its 161st Report on the Review of the Intellectual Property Rights Regime in India presented on 23 July 2021, recommended a separate category of rights for AI and AI-related innovations. Beyond these episodic engagements, neither the legislature nor the courts have authoritatively settled the doctrinal questions, and the discussion that follows seeks to map the terrain and consider whether the 1957 Act can be stretched to accommodate machine creativity.

Historical Development and Conceptual Framework

The English statutory ancestor of Indian copyright law, the Statute of Anne, 1710, framed copyright around the figure of the human author and her economic incentives. That orientation persisted through the Imperial Copyright Act, 1911, which was extended to British India by the Indian Copyright Act, 1914, and survived the post-independence overhaul effected by the Copyright Act, 1957. The Berne Convention for the Protection of Literary and Artistic Works, 1886, to which India acceded in 1928, similarly addresses itself to authors as natural persons and grants moral rights of attribution and integrity that presuppose a human personality. Even the Agreement on Trade-Related Aspects of Intellectual Property Rights, 1994, while pitched in more economic terms, builds on Berne and contains no explicit recognition of non-human authorship.

The conceptual core of the system is originality. In University of London Press Ltd v University Tutorial Press Ltd [1916] 2 Ch 601, Peterson J held that originality demands only that the work originate from the author and not be copied from another, the sweat-of-the-brow standard. That formulation was displaced in the United States by Feist Publications Inc v Rural Telephone Service Co 499 US 340 (1991), which required a modicum of creativity. The Indian Supreme Court in Eastern Book Company v D B Modak (2008) 1 SCC 1 charted an intermediate course, demanding the application of skill and judgment producing some minimal degree of creativity, while rejecting pure industrious effort. Each of these milestones assumes that the entity exercising skill, judgment or creativity is a natural person possessed of intellect, intention and aesthetic choice. Whether the same standards can be coherently applied to a stochastic model trained on terabytes of data is the question the present moment forces upon us.

Existing Legal Position in India

The Copyright Act, 1957, as amended most recently by the Copyright (Amendment) Act, 2012, defines the author in Section 2(d). For literary, dramatic, musical and artistic works the author is the person who creates the work; for cinematograph films and sound recordings, the producer; and, by virtue of Section 2(d)(vi) inserted by the Copyright (Amendment) Act, 1994, in relation to any literary, dramatic, musical or artistic work which is computer-generated, the author is the person who causes the work to be created. That provision was drafted with conventional computing in mind, where a programmer or user dictates the structure of the output. Its applicability to deep learning models, whose outputs emerge from statistical inferences over training data and exhibit a degree of unpredictability that approaches genuine novelty, is contested.

The text of Section 2(d) does not explicitly require the author to be a natural person, and the Copyright Office briefly accepted this elasticity when, in November 2020, it registered SURYAST listing RAGHAV as a co-author. A withdrawal notice followed on 25 November 2021, asking the applicant to clarify the legal status of the AI. The episode demonstrates the uncertainty within the registering authority itself. Section 13 of the Act lists the works in which copyright subsists and Section 17 designates the author as the first owner of copyright, subject to contractual variations. Moral rights under Section 57, including the right of paternity and the right against distortion, are premised on the author having a reputation and personality capable of being injured, attributes alien to a software system.

Indian courts have not yet ruled directly on AI authorship, but the contours of originality articulated in Eastern Book Company v D B Modak and reaffirmed in Tech Plus Media Private Ltd v Jyoti Janda 2014 (60) PTC 121 (Del) require a human exercise of skill and judgment that a machine cannot independently supply. The Indian Patent Office addressed an analogous question when it refused Stephen Thaler's application listing the DABUS system as inventor, holding that the Patents Act, 1970, contemplates only natural persons as inventors. The underlying logic, that intellectual property statutes presuppose human creativity, informs both regimes.

Comparative International Perspective

The United Kingdom occupies a singular position. Section 9(3) of the Copyright, Designs and Patents Act, 1988, provides that, for a literary, dramatic, musical or artistic work which is computer-generated, the author is the person by whom the arrangements necessary for the creation of the work are undertaken. Section 178 defines computer-generated to mean generated by a computer in circumstances such that there is no human author. This regime, replicated in jurisdictions such as Ireland, New Zealand and Hong Kong, is the closest analogue to a positive legislative answer. It has been criticised by commentators including Jyh-An Lee for failing to specify what arrangements are necessary and for fitting uneasily with the originality threshold articulated by the Court of Justice of the European Union in Infopaq International A/S v Danske Dagblades Forening Case C-5/08 [2009] ECR I-6569, which demands the author's own intellectual creation. The United Kingdom Intellectual Property Office consultation of 2021 considered repeal of Section 9(3) but ultimately deferred a decision.

In the United States, the Copyright Office has insisted on human authorship as a bedrock requirement. Compendium of US Copyright Office Practices, Third Edition, 2021, at paragraph 313.2 categorically excludes works produced by machines or mere mechanical processes that operate randomly or automatically without sufficient creative input. That position was tested in Thaler v Perlmutter, in which the United States District Court for the District of Columbia held on 18 August 2023 that human authorship is a bedrock requirement of copyright, an outcome affirmed by the Court of Appeals for the District of Columbia Circuit on 18 March 2025. The earlier judgment of the Ninth Circuit in Naruto v Slater 888 F 3d 418 (9th Cir 2018) had already foreshadowed this reasoning when it denied a non-human primate standing to assert copyright in the celebrated monkey selfies. Pending litigation in Andersen v Stability AI Ltd, filed in the Northern District of California on 13 January 2023, will further test whether the training of generative models on copyrighted images constitutes actionable infringement.

The European Union has approached the question primarily through the lens of training rather than authorship. Articles 3 and 4 of the Directive on Copyright in the Digital Single Market 2019/790 of 17 April 2019 introduce mandatory exceptions for text and data mining, the first restricted to research organisations and cultural heritage institutions, the second available to any user but subject to an opt-out by rightholders through machine-readable means. The Artificial Intelligence Act, Regulation (EU) 2024/1689 of 13 June 2024, imposes transparency obligations on providers of general-purpose AI models, including a duty to publish a sufficiently detailed summary of training content. Singapore, through Sections 243 and 244 of the Copyright Act 2021, has enacted a broader computational data analysis exception that permits both commercial and non-commercial use of lawfully accessed works for machine training, an approach designed to attract AI investment. Australia has been more conservative, with the Federal Court in Commissioner of Patents v Thaler [2022] FCAFC 62 of 13 April 2022 reversing an earlier decision and holding that only a natural person can be an inventor under the Patents Act, 1990 (Cth).

Major Challenges and Emerging Issues

The most immediate challenge is the authorship gap. If a generative model produces an image with minimal human direction beyond a brief prompt, neither the prompter nor the developer can comfortably claim the kind of creative contribution that Eastern Book Company demands. Yet to deny protection altogether risks consigning a growing share of cultural production to the public domain, with possible consequences for incentive structures in creative industries. The SURYAST registration and its subsequent suspension reveal precisely this institutional unease, since the Copyright Office could neither confidently reject the application nor coherently sustain it.

A second cluster of difficulties concerns infringement at the training stage. Large language and image models are trained on corpora that almost invariably include copyrighted works scraped from the open web. Whether such ingestion amounts to reproduction within Section 14 of the 1957 Act, and whether it can be defended under Section 52, is unresolved. Section 52(1)(a) applies only to private use, research, criticism or review, and the reporting of current events. The exhaustive Indian enumeration contrasts with the open-ended four-factor test in Section 107 of the United States Copyright Act, 1976, which has permitted American courts to accommodate transformative uses such as the search-index reproductions in Authors Guild v Google Inc 804 F 3d 202 (2d Cir 2015). Indian fair dealing offers no comparable elasticity, and a strict reading would place mass ingestion for commercial model training outside its scope.

A third issue concerns infringement at the output stage. Where a generative model produces text or imagery that substantially reproduces a training input, liability questions arise for the developer, deployer and end-user. The Delhi High Court in ANI Media Pvt Ltd v Open AI Inc, in which summons were issued in November 2024, will be among the first Indian forums to engage with these questions. A fourth difficulty lies under Section 57, since moral rights presuppose a personality capable of suffering reputational harm and extending them to algorithmic agents is conceptually awkward. A fifth is the cross-border dimension, as training datasets are assembled in one jurisdiction, models hosted in another and outputs deployed globally.

Critical Analysis

Three positions have crystallised in the academic literature. The first, defended by scholars such as Ryan Abbott, urges the recognition of AI as author or co-author so that the economic incentive of copyright continues to channel investment into creative technologies. The second, articulated by Annemarie Bridy and Jane Ginsburg, insists that human authorship is non-negotiable and that AI-generated material lacking human creative input belongs in the public domain. The third, advanced by Daniel Gervais and others, advocates a sui generis right of shorter duration and narrower scope for autonomously generated works, modelled loosely on the European database right under Directive 96/9/EC of 11 March 1996.

The Indian statutory architecture, read carefully, leans towards the second position but with a doctrinal opening towards the first. Section 2(d)(vi) does not require the person who causes the work to be created to have exercised personal skill and judgment; it suffices that the person causes the work, which a user supplying prompts arguably does. Yet Eastern Book Company demands a modicum of creativity, and the originality threshold cannot be met by causation alone. The tension between Section 2(d)(vi) and the judicial originality doctrine has not been resolved, and a future court confronted with an AI-generated work will be forced to choose between a literal reading of the statute and a doctrinally consistent application of Eastern Book Company. The Standing Committee in Report No. 161 of July 2021 recognised this dilemma and recommended the creation of a separate category of rights, a recommendation that has not yet found legislative expression.

The training-data question is equally fraught. Section 52 of the 1957 Act is a closed list, and extending it by judicial interpretation to encompass commercial machine training would strain the fair dealing doctrine to a breaking point. The Singaporean solution under Sections 243 and 244 of the Copyright Act, 2021, while pragmatically attractive, was the product of a deliberate legislative choice, not interpretive enlargement. The European compromise in Article 4 of the Directive 2019/790 preserves right holder autonomy through an opt-out mechanism, an approach that may sit more comfortably with the Indian constitutional emphasis on the property rights of authors recognised in Article 300A of the Constitution and the broader fundamental rights jurisprudence developed in K S Puttaswamy v Union of India (2017) 10 SCC 1 in relation to informational autonomy.

A distributive concern also bears mention. The largest beneficiaries of an expansive reading of Section 2(d)(vi) would be the corporations that train and deploy generative models, while the costs would fall on individual authors whose works form the unconsented training corpus. A coherent Indian response must attend not only to the doctrinal question of authorship but also to the political economy of generative AI, ensuring that any extension of protection is accompanied by mechanisms for fair remuneration of upstream creators.

Suggestions and Recommendations

A targeted legislative response, rather than a complete reconstruction of the 1957 Act, appears most appropriate. Parliament could clarify Section 2(d)(vi) by inserting an explanation requiring that the person who causes the work to be created must have exercised non-trivial creative direction over the output. Such an amendment would align statutory text with the originality doctrine in Eastern Book Company and forestall opportunistic claims based on rudimentary prompts. For outputs generated with no meaningful human creative contribution, a sui generis right of limited duration, perhaps fifteen or twenty years, could be considered, with reduced moral rights and explicit recognition that the human investor in the AI system holds the economic rights. This approach echoes the recommendation of the Parliamentary Standing Committee in its 161st Report of 2021.

On training data, India should reform Section 52 to introduce a calibrated text and data mining exception. A non-commercial research exception of the kind found in Article 3 of Directive 2019/790 would be relatively uncontroversial. A commercial exception modelled on Article 4, with a transparent opt-out mechanism for rightholders, would balance India's ambition to host AI development against the legitimate interests of authors and publishers. Transparency obligations on developers, comparable to those in the European Artificial Intelligence Act, 2024, would enable rightholders to detect infringement and exercise opt-outs meaningfully. The Copyright Rules, 1958, could be amended to require disclosure of training data summaries at the point of registration of any AI-generated work.

Forums confronted with cases such as ANI Media Pvt Ltd v Open AI Inc should resist legislating from the bench and enforce the existing text of Section 52 strictly, leaving policy innovation to Parliament. The Copyright Office should issue formal guidelines requiring disclosure of the nature and extent of human creative contribution at the point of registration, replacing the ad hoc approach that produced the SURYAST controversy. India should also participate actively in the World Intellectual Property Organization Conversation on Intellectual Property and Frontier Technologies, initiated in 2019, to shape an international consensus rather than confront the issue in isolation.

Conclusion

Indian copyright law stands at a doctrinal crossroads. The Copyright Act, 1957, drafted for an era of human creativity, can absorb modest pressure from generative artificial intelligence through interpretive flexibility, but cannot withstand the sustained stress that fully autonomous machine creation will impose. The brief recognition and withdrawal of the SURYAST registration, the unresolved recommendation of the Standing Committee in Report No. 161 of 2021, and the litigation now reaching Indian courts together indicate that the moment for legislative engagement has arrived. No single jurisdiction has yet found a satisfactory solution. The United Kingdom regime under Section 9(3) of the 1988 Act is doctrinally uneasy, the American insistence on human authorship leaves machine outputs in a normative vacuum, the European compromise is administratively complex, and the Singaporean exception favours developers over upstream creators. India has the opportunity to fashion a measured response that calibrates Section 2(d)(vi), introduces a balanced text and data mining exception in Section 52, and contemplates a sui generis right for genuinely autonomous output. What is required is not the abandonment of the 1957 Act but its thoughtful evolution to meet a technology its drafters could not have foreseen.

Bibliography

Primary Sources

Statutes and Legislative Instruments

Copyright Act 1957 (India)

Copyright (Amendment) Act 1994 (India)

Copyright (Amendment) Act 2012 (India)

Copyright Rules 1958 (India)

Patents Act 1970 (India)

Constitution of India 1950

Copyright, Designs and Patents Act 1988 (UK)

Copyright Act 1976 (US), 17 USC

Copyright Act 2021 (Singapore)

Patents Act 1990 (Cth) (Australia)

Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market [2019] OJ L130/92

Directive 96/9/EC of the European Parliament and of the Council of 11 March 1996 on the Legal Protection of Databases [1996] OJ L77/20

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 (Artificial Intelligence Act) [2024] OJ L1689

Berne Convention for the Protection of Literary and Artistic Works 1886 (as revised at Paris 1971)

Agreement on Trade-Related Aspects of Intellectual Property Rights 1994, 1869 UNTS 299

Cases

Eastern Book Company v D B Modak (2008) 1 SCC 1

Tech Plus Media Private Ltd v Jyoti Janda 2014 (60) PTC 121 (Del)

ANI Media Pvt Ltd v Open AI Inc, CS(COMM) 1028/2024 (Del HC)

University of London Press Ltd v University Tutorial Press Ltd [1916] 2 Ch 601

Infopaq International A/S v Danske Dagblades Forening (Case C-5/08) [2009] ECR I-6569

Feist Publications Inc v Rural Telephone Service Co 499 US 340 (1991)

Authors Guild v Google Inc 804 F 3d 202 (2d Cir 2015)

Naruto v Slater 888 F 3d 418 (9th Cir 2018)

Thaler v Perlmutter 687 F Supp 3d 140 (DDC 2023), aff'd No 23-5233 (DC Cir 18 March 2025)

Andersen v Stability AI Ltd, No 3:23-cv-00201 (ND Cal, filed 13 January 2023)

Commissioner of Patents v Thaler [2022] FCAFC 62

Government and Institutional Reports

Department-Related Parliamentary Standing Committee on Commerce, Review of the Intellectual Property Rights Regime in India (Report No 161, Rajya Sabha Secretariat, 23 July 2021)

United States Copyright Office, Compendium of US Copyright Office Practices (3rd edn, 2021)

United States Copyright Office Review Board, Second Request for Reconsideration for Refusal to Register SURYAST (11 December 2023)

UK Intellectual Property Office, Artificial Intelligence and Intellectual Property: Copyright and Patents — Government Response to Consultation (28 June 2022)

World Intellectual Property Organization, WIPO Conversation on Intellectual Property and Frontier Technologies (Geneva, ongoing since 2019)

Secondary Sources

Books

Abbott R, The Reasonable Robot: Artificial Intelligence and the Law (Cambridge University Press 2020)

Gervais D, (Re)structuring Copyright: A Comprehensive Path to International Copyright Reform (Edward Elgar 2017)

Lee J-A, Hilty R and Liu K-C (eds), Artificial Intelligence and Intellectual Property (Oxford University Press 2021)

Narayanan P, Law of Copyright and Industrial Designs (4th edn, Eastern Law House 2017)

Journal Articles

Bridy A, 'Coding Creativity: Copyright and the Artificially Intelligent Author' (2012) 5 Stanford Technology Law Review 1

Ginsburg J C and Budiardjo L A, 'Authors and Machines' (2019) 34 Berkeley Technology Law Journal 343

Lee J-A, 'Dealing with AI-Generated Works: Lessons from the CDPA Section 9(3)' (2024) 19 Journal of Intellectual Property Law and Practice 43

Geiger C, Frosio G and Bulayenko O, 'The Exception for Text and Data Mining in the Proposed Directive on Copyright in the Digital Single Market' (2018) Centre for International Intellectual Property Studies Research Paper No 2018-02

Samuelson P, 'Allocating Ownership Rights in Computer-Generated Works' (1986) 47 University of Pittsburgh Law Review 1185

Online and News Sources

Reddy P and Chowdhury T, 'AI Art and Indian Copyright Registration' (SpicyIP, 14 October 2022) <https://spicyip.com/2022/10/ai-art-and-indian-copyright-registration.html>

Patnaik R, 'Exclusive: Indian Copyright Office Issues Withdrawal Notice to AI Co-Author' (Managing IP, December 2021) <https://www.managingip.com>

PRS Legislative Research, Summary of Standing Committee Report on Review of the Intellectual Property Rights Regime in India (PRS India, 2021) <https://prsindia.org>