Copyright Protection for AI-Generated Works in India

ABSTRACT

The emergence of generative artificial intelligence systems capable of autonomously producing literary, artistic, musical, and computational works has exposed a fundamental structural lacuna within India’s copyright architecture. This article examines whether the Copyright Act, 1957, as presently constituted, extends protection to works generated by AI systems without direct human creative intervention, analyses the definitional constraints of section 2(d)(vi), and evaluates the “skill and judgment” standard articulated in Eastern Book Company v D B Modak against the technological realities of autonomous creative AI. Through doctrinal analysis, comparative jurisprudence across the United Kingdom, United States, European Union, and China, and constitutional examination under Articles 14, 19, and 21, this article concludes that India’s anthropocentric copyright framework is structurally inadequate to address AI-generated content and that targeted legislative reform — creating a sui generis protection regime with calibrated term, mandatory disclosure, and equitable remuneration — is both constitutionally permissible and commercially necessary. The article sets out eight findings and ten legislative recommendations directed at Parliament, the Copyright Office, and the Ministry of Electronics and Information Technology.

Keywords: artificial intelligence; copyright; authorship; originality; sui generis; section 2(d)(vi); generative AI; Copyright Act 1957; India; legislative reform; comparative law; constitutional law

I. INTRODUCTION

The twenty-first century has witnessed the emergence of generative artificial intelligence systems of unprecedented creative capability. Large language models such as OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude produce publishable literary prose, legal memoranda, and journalistic content at scale; image-synthesis models such as Midjourney, DALL-E, and Stable Diffusion generate fine-art-quality visuals from textual prompts; music-generation platforms compose original compositions; and code-generation tools write functional software deployed in commercial products. These outputs are commercially deployed across industries ranging from advertising and publishing to healthcare and financial services, generating substantial economic value and posing foundational challenges to every legal system premised upon human creative authorship.[1]

The challenge to copyright law is structural, not merely interpretive. Copyright regimes across the world have, since the Berne Convention of 1886, premised the subsistence of copyright upon a human author whose intellectual and creative personality is expressed in the protected work.[2] The generative AI system, by contrast, operates through stochastic mathematical processes upon massive training corpora to produce outputs over which no single human exercises granular creative control in real time. When a user inputs a textual prompt and receives a richly detailed creative output, it is genuinely contested whether any human being has exercised the kind of creative judgment that copyright law has historically required as a condition of protection.

In India, this uncertainty is legally acute. The Copyright Act, 1957, rests upon an anthropocentric authorship paradigm embedded in its definitional structure. Section 2(d), even after its amendment in 1994 to introduce clause (vi) addressing computer-generated works,[3] continues to presuppose a human agent who “causes the work to be made.” The Supreme Court’s endorsement in Eastern Book Company v D B Modak[4] of the “skill and judgment” standard provides no ready doctrinal answer when the skill is exercised by a transformer-based neural network and the human contribution is confined to the specification of a textual prompt. The absence of legislative guidance has left the Indian creative economy, technology sector, and legal community operating in a condition of doctrinal uncertainty that requires urgent resolution.

The commercial stakes are formidable. India’s generative AI market is projected to reach USD 17 billion by 2030, with AI-generated content constituting a growing share of digital economy output. Legal uncertainty regarding ownership of AI-generated works impedes investment, creates contractual risk, and disadvantages Indian developers and creative industries in global markets. The absence of a legislative response — at a moment when the United Kingdom, United States, European Union, and China are each actively developing regulatory positions — risks leaving India as a rule-taker rather than a rule-maker in the global governance of AI-generated intellectual property.

This article proceeds in fifteen parts. Parts II to IV frame the research questions, objectives, and methodology; Part V surveys the historical evolution of copyright law in relation to AI; Part VI examines the Indian copyright framework in detail; Part VII analyses relevant Indian judicial decisions with full case analysis; Part VIII provides comparative analysis across four jurisdictions; Part IX identifies major legal challenges; Part X examines constitutional dimensions; Part XI explores economic and policy implications; Part XII offers critical evaluation; Parts XIII and XIV set out findings and recommendations; and Part XV concludes.

II. RESEARCH QUESTIONS

This article is directed at answering the following five research questions:

1. Does section 2(d)(vi) of the Copyright Act, 1957, on a proper textual and purposive construction, extend to works generated autonomously by AI systems without qualifying human creative contribution?

2. What minimum standard of human creative contribution, if any, is constitutionally required for copyright subsistence in India, and does the Constitution permit Parliament to protect AI-generated works irrespective of human authorship?

3. How have the United Kingdom, United States, European Union, and China resolved the AI authorship question, and what lessons does comparative experience offer for Indian legislative design?

4. What regulatory and institutional mechanisms are necessary to operationalise a reformed copyright framework for AI-generated works in India, including disclosure, registration, and licensing infrastructure?

5. Is a sui generis protection regime for AI-generated works — distinct from conventional copyright — constitutionally permissible and commercially desirable in the Indian context?

III. OBJECTIVES OF STUDY

1.  To analyse the textual and purposive scope of section 2(d)(vi) of the Copyright Act, 1957, in light of the legislative history of the 1994 Amendment and contemporary AI capabilities.

2.  To evaluate the doctrinal adequacy of the “skill and judgment” standard in Eastern Book Company v D B Modak when applied to AI-generated content.

3.  To examine the constitutional dimensions of AI copyright protection under Articles 14,19(1)(a), 19(1)(g), and 21 of the Constitution of India.

4.  To survey comparative legislative and judicial approaches across the UK, USA, EU, andChina.

5.  To identify the principal legal challenges arising from generative AI, including theauthorship problem, training data infringement, moral rights, and public domain concerns.

6.  To assess the economic and policy implications of different regulatory approaches for India’screative economy, technology sector, and innovation ecosystem.

7.  To formulate concrete, constitutionally grounded recommendations for legislative reform.

IV. RESEARCH METHODOLOGY

This study employs a doctrinal research methodology complemented by analytical and comparative approaches. Doctrinal analysis examines the statutory text of the Copyright Act, 1957 — particularly sections 2(d), 13, and 17 — in light of the principles of statutory interpretation established by the Supreme Court of India, including literal, golden, and purposive construction. The legislative history of the 1994 Amendment, including the Statement of Objects and Reasons, is examined to identify the intended scope of section 2(d)(vi) in relation to the TRIPS Agreement.[5]

Analytical methodology is employed in Part XII to critically evaluate the policy choices embedded in competing regulatory models. Comparative methodology guides Parts VIII through XI: the UK Copyright, Designs and Patents Act 1988, US Copyright Office guidelines, EU AI Act 2024, and Chinese regulatory instruments are examined as tested regulatory alternatives. Primary sources include statutes, constitutional provisions, judicial decisions, Law Commission reports, and international instruments. Secondary sources include peer-reviewed scholarship from Guadamuz, Ginsburg, Samuelson, and Nimmer on AI and intellectual property law.

V. EVOLUTION OF COPYRIGHT LAW AND ARTIFICIAL INTELLIGENCE

Copyright protection originated in eighteenth-century England with the Statute of Anne, 1710, which vested rights in authors rather than publishers for the first time. The philosophical foundations of copyright — whether grounded in Lockean labour theory, Kantian personality theory, or Utilitarian incentive theory — uniformly presuppose a human creative subject whose intellectual contribution deserves legal protection. The Berne Convention of 1886,[6] which established the framework of international copyright, defined protected works in terms of “literary and artistic productions” — an inherently anthropocentric category. Subsequent international instruments, including the TRIPS Agreement of 1994[7] and the WIPO Copyright Treaty of 1996,[8] have maintained this foundational assumption without substantive revision.

The first serious challenge to the human-authorship premise arose with the development of computer-generated works in the 1980s. Rule-based programs capable of generating musical compositions, legal documents, and news reports from structured data prompted legislative responses. The United Kingdom, in section 9(3) of the Copyright, Designs and Patents Act 1988,[9] attributed authorship of computer-generated works to the person who “undertakes the arrangements necessary for the creation of the work” — a provision that India subsequently adopted in modified form in section 2(d)(vi). These legislative solutions were adequate for rule-based, deterministic software but are conceptually strained when applied to self-learning generative AI systems of the present era.

The development of machine learning, deep learning, and transformer architecture — from the 2012 AlexNet breakthrough through the 2017 “Attention Is All You Need” paper and the 2022 public deployment of large-scale language models — represents a qualitative shift in AI capability. Where rule-based software executes explicit programmer-defined instructions to produce deterministic outputs, generative AI systems extract statistical patterns from massive training corpora and produce novel outputs through a probabilistic process that their own developers cannot fully predict or control. The human creative contribution in AI-generated content has become so attenuated that existing statutory concepts — authorship, originality, skill, judgment — can no longer be applied without doctrinal distortion.

VI. INDIAN COPYRIGHT FRAMEWORK

A. Section 2(d): The Authorship Definition and Section 2(d)(vi)

Section 2(d) of the Copyright Act, 1957, defines “author” with respect to different categories of works. Critically, section 2(d)(vi) provides that in relation to any literary, dramatic, musical, or artistic work which is computer-generated, the author shall be taken to be “the person by whom the arrangements necessary for the creation of the work are undertaken.”[10] This is the only textual basis in Indian law for attributing authorship to AI-generated content. The provision was modelled on CDPA 1988 s 9(3)[11] and was intended to address deterministic, rule-based programs — not autonomous generative AI systems. Parliament’s purpose in 1994 was to attribute copyright in computer-generated works to the human who programmed the computer, not to vest rights in systems whose outputs are the product of autonomous machine learning processes over which no programmer exercises granular creative control.

B. Section 13: Subsistence of Copyright and the Originality Requirement

Section 13(1) of the Copyright Act vests copyright in “original literary, dramatic, musical and artistic works.”[12] The Act does not define “originality,” leaving its content to judicial development. The Supreme Court, culminating in Eastern Book Company v D B Modak, has settled that originality in Indian law requires the work to reflect the author’s own intellectual creation involving the exercise of skill and judgment — rejecting the lower “sweat of the brow” doctrine while not requiring a threshold of artistic quality. Applied to AI-generated works, this standard produces a doctrinal paradox: the AI system exercises the “skill” while the human operator contributes at most a directional prompt -arguably too thin a creative contribution to satisfy the statutory requirement under the current standard.

C. Section 17: First Ownership and the Multi-Party Attribution Problem

Section 17 of the Copyright Act provides that the author of a work is the first owner of copyright therein,[13] subject to exceptions for works made in the course of employment. Applied to AI-generated works, section 17 produces a fundamental indeterminacy: if section 2(d)(vi) attributes authorship to the person who makes “arrangements necessary” for the work’s creation, does this mean the model developer (who created the underlying architecture), the model trainer (who curated the training data), the platform operator (who deployed the model), or the end user (who input the specific prompt)? The statute provides no principled basis for resolution. Multi-party disputes over ownership of AI-generated content will inevitably arise, imposing costs on all stakeholders and creating litigation uncertainty that legislative intervention can prevent.

VII. JUDICIAL ANALYSIS

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

Facts: The appellant published annotated Supreme Court law reports incorporating editorial notes, headnotes, cross-references, and indexing. The respondent reproduced these materials without permission, claiming the underlying judgments were public-domain documents not susceptible to copyright.

Issue: Whether the editorial contribution to the compilation of Supreme Court judgments satisfied the originality standard for copyright subsistence under the Copyright Act, 1957.

Judgment: A two-judge bench of the Supreme Court held that copyright subsists only where the work reflects the author’s own “intellectual creation” involving “independent skill, labour, and judgment,”[14] explicitly rejecting the “sweat of the brow” doctrine. The Court adopted a standard consonant with the Canadian Supreme Court’s approach in CCH Canadian Ltd v Law Society of Upper Canada.[15] The editorial notes and headnotes satisfied this standard; verbatim reproduction of judgments did not.

Principle and Relevance to AI: The Eastern Book Company standard presupposes a sentient human author exercising independent creative judgment. When a language model produces a legal memorandum, the “skill” is that of the model’s architecture and training data; the “judgment” is the statistical output of billions of parameter weights. The human prompter contributes at most a directional input — arguably insufficient to satisfy the current statutory requirement. This doctrinal paradox is the primary engine driving the call for legislative reform of the Copyright Act.

B. Navigators Logistics Ltd v Kashif Qureshi (2018 SCC OnLine Del 11303)

Facts: The plaintiff claimed copyright in standard business contracts and commercial documents prepared by its employees for use in company operations.

Issue: Whether copyright subsisted in standard-form commercial documents prepared by employees and whether their reproduction by a former employee constituted infringement.

Judgment and Relevance: The Delhi High Court affirmed that copyright requires human authorship and independent intellectual contribution,[16] confirming the Eastern Book Company standard in the context of commercial documents. While the case did not directly address AI-generated documents, the affirmation of the human authorship requirement provides the foundational constraint against which any AI copyright claim must be assessed under current Indian law.

C. R G Anand v Deluxe Films AIR 1978 SC 1613

Facts: The appellant claimed that the defendant’s cinematograph film was substantially copied from his dramatic work, raising the question of the threshold of copyright infringement.

Issue: What constitutes copyright infringement in creative works, and where does the idea-expression divide lie?

Judgment and Relevance: The Supreme Court held that an “idea” per se is not copyrightable; copyright protects only the original expression of an idea.[17] This idea-expression dichotomy has profound implications for AI copyright: a text prompt specifying an idea or concept does not itself attract copyright, nor may it generate copyright in the AI output if that output is the product of machine learning rather than human creative expression. The doctrine additionally limits the liability of AI operators for generating outputs that are stylistically similar to existing works but do not reproduce their specific expression.

D. Indian Performing Rights Society Ltd v Eastern Indian Motion Pictures Association AIR 1977 SC 1443

The Supreme Court confirmed that copyright vests in the creator of an original work and that assignment and licensing require specific statutory compliance.[18] In the AI context, this ruling’s emphasis on clear ownership attribution underscores the legislative imperative: without a clear statutory rule allocating copyright or a sui generis right in AI-generated works, the ownership question will generate protracted litigation that imposes costs on AI developers, platforms, and creative industries alike.

VIII. COMPARATIVE JURISDICTION ANALYSIS

A. United Kingdom: The CDPA 1988 “Arrangements” Framework

The United Kingdom is the only major jurisdiction with explicit statutory provision for computer-generated works. Section 9(3) of the Copyright, Designs and Patents Act 1988 attributes authorship to “the person by whom the arrangements necessary for the creation of the work are undertaken.”[19] This provision — on which section 2(d)(vi) of the Indian Act was modelled — has been criticised by the UK Intellectual Property Office and by leading copyright scholars for its circularity: it does not resolve who qualifies as the “arranger” when an autonomous AI system makes all creative decisions. The UK IPO, in its 2022 consultation on AI and Intellectual Property, considered abolishing this provision in favour of a fresh framework but reached no final legislative recommendation, leaving the UK’s law as uncertain as India’s despite three decades of experience with the provision.

B. United States: The Human Authorship Doctrine and Thaler v Perlmutter

The United States Copyright Office has consistently denied copyright registration to AI-generated works. In Thaler v Perlmutter,[20] the US District Court for the District of Columbia upheld the Copyright Office’s refusal to register an AI-generated image, holding that copyright as a constitutional matter requires human authorship under Article I, Section 8, Clause 8 of the US Constitution. The US Copyright Office, in its 2024 Part II Guidance on Copyright and Artificial Intelligence,[21] confirmed that AI-generated content without sufficient human creative contribution is not copyrightable, while AI-assisted content — where a human author makes qualifying creative choices — may attract protection for the human-authored elements. The Feist standard[22] of minimal creativity by a human author, combined with the constitutional requirement, creates a two-tier barrier that is higher than what the Indian framework need necessarily maintain. India’s Parliament is not constitutionally required to premise copyright exclusively on human authorship, giving India greater legislative flexibility than the United States.

C. European Union: AI Act Transparency and the Copyright Directive

The European Union’s Artificial Intelligence Act of 2024[23] imposes transparency obligations on providers of general-purpose AI models, including requirements to disclose when content is AI-generated and to comply with EU copyright law when using copyrighted training data. The AI Act does not directly address ownership of AI-generated outputs, leaving this question to Member State copyright law. The EU Copyright Directive’s Article 4 text-and-data-mining exception provides a partial model for addressing training data infringement, creating an opt-out right for rights-holders while enabling AI training at scale. The EU’s experience demonstrates that sector-specific transparency regulation can proceed without resolving the authorship question, providing India with a partial legislative model for a phased reform approach.

D. China: The Most Proactive Comparative Framework

China has adopted the most proactive judicial approach to AI-generated content. The Beijing Internet Court, in Tencent v Shanghai Yingxun Technology Co (2019),[24] held that AI-generated content could attract copyright where it was generated through a sufficiently complex process of human design, parameter setting, and creative direction — even if the final output was produced by an algorithm. This “human-design” approach offers India a middle path: protecting AI-generated content where demonstrable human creative direction can be established, without requiring the granular, moment-by-moment control that the Eastern Book Company standard may imply. China’s regulatory instruments for deepfake content and synthetic media further provide models for disclosure and labelling obligations that India’s proposed sui generis framework could incorporate.

IX. MAJOR LEGAL CHALLENGES

Authorship Problem: Current copyright doctrine has no principled basis for identifying the

“author” of an AI-generated work among the developer, trainer, operator, and user. Section 2(d)(vi)’s “arrangements necessary” language does not resolve this multi-party attribution problem, and the courts have not yet addressed it in the AI context.

Ownership Ambiguity: Even if authorship can be attributed, the chain of copyright ownership in AI-generated works is complex. Developer-operator licensing agreements, user terms of service, and employment contracts all potentially affect who owns the output, creating multi-party disputes in the absence of clear statutory rules.

Originality Standards: The Eastern Book Company “skill and judgment” standard likely excludes most AI-generated works from copyright protection. A reformed originality standard recognising a spectrum from human-only to AI-only creation is necessary to serve the copyright system’s incentive functions in the generative AI era.

Training Data Infringement: Generative AI systems are trained on massive corpora including copyrighted works without licence. Indian copyright law contains no clear text-and-data-mining exception, leaving AI developers exposed to copyright liability in respect of training activities — a significant chilling effect on AI development in India.

Moral Rights: Section 57 of the Copyright Act confers on authors the right of attribution and the right of integrity. If AI-generated works attract copyright, it is unclear what moral rights should attach: an AI system cannot be defamed, and no human author whose personality is expressed in the work can be identified for moral rights purposes.

Deepfakes and Derivative Works: AI systems generate outputs that may be derivative of copyrighted training works. The boundary between permissible stylistic influence and actionable reproduction requires statutory clarification in the AI context. Real-life harms from AI-generated impersonations of celebrities and politicians further demand targeted legislative response.

Public Domain Paradox: If AI-generated works enter the public domain immediately upon creation, commercially valuable content is available for free appropriation by all -disincentivising AI development. Conversely, vesting copyright in AI operators risks monopolisation of creative output by a small number of powerful technology companies, raising competition and access-to-knowledge concerns that Parliament must carefully balance.

X. CONSTITUTIONAL PERSPECTIVE

Article 19(1)(a) of the Constitution protects freedom of speech and expression, encompassing the right to receive and access information.[25] The vesting of copyright in AI operators over large volumes of AI-generated content could restrict public access to information and constrain the creative commons, implicating the Article 19(1)(a) rights of citizens. Any copyright protection for AI-generated works must include robust limitations and exceptions — including a fair dealing provision for education, research, and commentary — to protect these rights.

Article 19(1)(g) protects the right to practise any profession and carry on any occupation, trade, or business,[26] which encompasses the exploitation of intellectual property rights in commercial contexts. The Supreme Court in Internet and Mobile Association of India v Reserve Bank of India[27] affirmed that disproportionate restrictions on economically valuable activities require strong justification. A complete denial of copyright to AI-generated works would restrict AI developers’ ability to recoup investments through IP rights, potentially engaging Article 19(1)(g) proportionality review under Article 19(6).

Article 21, as expansively interpreted through the Maneka Gandhi[28] and Puttaswamy[29] framework, encompasses the right to privacy and informational self-determination. The use of individuals’ personal data and creative works in AI training corpora without consent implicates the privacy rights of the individuals whose data is processed. Any reformed copyright framework must integrate DPDPA 2023 compliance requirements for AI training data collection and use, as well as the Puttaswamy proportionality test for any surveillance-enabling dimensions of AI content regulation.

XI. ECONOMIC AND POLICY IMPLICATIONS

India’s startup ecosystem — comprising over 100,000 startups as of 2024, many in AI and technology sectors — stands to benefit substantially from clear copyright rules for AI-generated content. Legal certainty enables investment, facilitates IP-backed financing, and allows startups to commercialise AI-generated products without litigation risk.[30] Conversely, the absence of clarity creates a first-mover disadvantage vis-a-vis US and EU competitors operating within more defined — if still evolving — legal frameworks. The IndiaAI Mission’s USD 1.25 billion investment in computing infrastructure underscores the national strategic importance of clear IP rules for AI-generated content.

India’s creative industries — Bollywood, the music industry, the publishing sector — face significant disruption from AI systems that can replicate styles, voices, and visual aesthetics without infringement liability under current law. A reformed framework must balance the interests of human creators, who deserve protection from displacement, against the interests of innovators and consumers who benefit from the democratisation of creative production that AI enables. A compulsory licensing fund financed by levies on AI-generated content platforms could provide equitable remuneration for displaced human creators.

The Digital Personal Data Protection Act, 2023[31] intersects with AI copyright reform: training data comprising personal data must comply with the DPDPA’s consent requirements, and the DPDPA’s data minimisation obligations may constrain the scope of permissible training data collection and use. A comprehensive AI governance framework for India must integrate copyright reform with DPDPA compliance, creating a coherent legal architecture for the AI-generated content economy.

XII. CRITICAL EVALUATION

The prevailing doctrinal position in India — which implicitly denies copyright to AI-generated works produced without sufficient human creative contribution — generates a “public domain paradox.” Commercially valuable content produced by AI systems enters the public domain immediately upon creation, available for free appropriation by any person. This result disincentivises AI development and investment, advantages jurisdictions that provide protection, and may ultimately harm Indian creators who would benefit from a robust domestic AI industry capable of generating export revenues and creating skilled employment.

The argument for a sui generis regime — analogous to the EU database right created by the Database

Directive of 1996 — is persuasive on economic grounds and is constitutionally available to Parliament under Entry 49 of List I of the Seventh Schedule, which confers a broad mandate over “patents, inventions, designs, copyright and trademarks.” India has the constitutional flexibility to design a protection regime calibrated to the AI context without being constrained by the historical assumptions of the 1957 Act.

The strongest argument against AI copyright is the democratic concern: vesting copyright in AI operators would effectively gift monopoly rights over large volumes of content to a small number of powerful technology companies, entrenching concentration in the creative economy. This concern can be addressed through careful legislative design: a reduced protection term of fifteen years, compulsory licensing provisions, and equitable remuneration mechanisms can limit the monopolistic effects while still providing sufficient commercial incentive for AI development and deployment.

The human creative contribution spectrum provides the most analytically coherent framework for resolving the AI copyright question. At one extreme, purely autonomous AI generation without any human creative input deserves no copyright but may attract a limited sui generis right. At the other extreme, AI-assisted human creation — where a human author uses AI as a sophisticated tool while retaining creative control and exercising independent judgment — deserves full copyright protection as the expression of human creativity. Between these poles, a graduated regime calibrated to the degree of human creative contribution provides the most nuanced and constitutionally defensible response.

XIII. FINDINGS

Finding 1: Section 2(d)(vi) of the Copyright Act, 1957, does not, on a proper purposive construction, extend to works generated autonomously by generative AI systems without qualifying human creative contribution. The provision was enacted to address rule-based, deterministic software and its purposive scope does not encompass self-learning generative models whose outputs are the product of stochastic probabilistic processes.

Finding 2: The “skill and judgment” standard in Eastern Book Company v D B Modak presupposes a sentient human author exercising independent creative judgment. Applied to AI-generated outputs, the standard produces a doctrinal paradox: the AI system exercises the “skill” while the human operator contributes at most a directional prompt insufficient to satisfy the statutory requirement as currently interpreted.

Finding 3: Comparative jurisdictions have not converged on a uniform solution. The UK CDPA 1988 s 9(3) approach is acknowledged as inadequate for autonomous AI; the US constitutional human authorship requirement forecloses statutory flexibility; the EU is developing transparency obligations without settling ownership; and China has adopted the most permissive judicial approach. India has an opportunity to adopt a contextually calibrated, constitutionally grounded framework that learns from each jurisdiction’s experience.

Finding 4: The current legal position creates a public domain paradox that disincentivises AI development in India. Commercially valuable AI-generated content enters the public domain immediately upon creation, disadvantaging Indian AI developers relative to competitors in jurisdictions that provide protection and driving investment away from India’s generative AI sector.

Finding 5: The constitutional framework under Articles 14, 19, and 21 permits — but does not require — Parliament to extend copyright or a sui generis right to AI-generated works. The constitutional constraint is not a bar to legislative reform but imposes proportionality requirements: any protection regime must be non-arbitrary, not disproportionately restrict expression or access to knowledge, and be accompanied by adequate procedural safeguards.

Finding 6: The training data infringement question is as legally significant as the authorship question. The absence of a clear text-and-data-mining exception in Indian copyright law exposes AI developers to infringement liability in respect of training activities, creating uncertainty that chills AI development and investment in India.

Finding 7: Moral rights under section 57 of the Copyright Act cannot coherently attach to AI-generated works in the absence of a human author. Any legislative regime for AI-generated works should either exclude moral rights entirely or vest them in a human person nominated for the purpose by the operator of the AI system.

Finding 8: The intersection of the DPDPA 2023 and AI training data creates a compliance challenge for Indian AI developers that must be addressed through coordinated regulatory guidance from the Copyright Office, the Data Protection Board of India, and the Ministry of Electronics and Information Technology, creating a coherent cross-regulatory framework for AI-generated content.

 

XIV. RECOMMENDATIONS

 

Recommendation 1 — Dedicated AI Copyright Chapter: Parliament should amend the Copyright Act, 1957, to insert a new Chapter IIA governing “AI-generated works,” providing a comprehensive statutory framework with clear definitions, ownership rules, limitations, and exceptions tailored specifically to the AI content generation context.

Recommendation 2 — Sui Generis Protection with Reduced Term: AI-generated works produced without sufficient human creative contribution should attract a sui generis right — not copyright — vested in the operator of the AI system, for a term of fifteen years from publication, after which the work enters the public domain. This provides a commercially meaningful incentive period while preventing perpetual monopolisation.

Recommendation 3 — Mandatory AI Disclosure and Labelling: Any AI-generated work commercially published or distributed in India should carry a mandatory disclosure identifying the AI system used, the operator, and the date of generation. The Copyright Office should develop a standardised disclosure format implemented across all digital publishing platforms.

Recommendation 4 — AI Content Registry: The Copyright Office should establish a voluntary AI Content Registry, allowing operators to register AI-generated works and creating a public record that facilitates licensing, attribution, and enforcement. Registration should trigger a rebuttable presumption of the registered operator’s sui generis right.

Recommendation 5 — Text-and-Data-Mining Exception: Parliament should insert a clear text-and-data-mining exception into the Copyright Act, permitting the use of copyrighted works for AI training purposes subject to conditions of transparency and rights-holder notification, analogous to Article 4 of the EU Copyright Directive.

Recommendation 6 — Transparency and Algorithmic Accountability: Providers of AI systems that generate commercially deployed content should be required to disclose the nature of training data used, the model architecture, and quality assurance processes applied to outputs, enabling rights-holders and regulators to assess compliance.

Recommendation 7 — Graduated Copyright Framework: The Copyright Office should issue guidelines establishing a spectrum from “fully human-authored” through “AI-assisted” to “fully AI-generated” works, with different levels of protection and distinct ownership rules applying to each category along the creative contribution spectrum.

Recommendation 8 — Fair Compensation Mechanism for Human Creators: Parliament should establish a compulsory licensing fund — financed by levies on AI-generated content platforms -that distributes royalties to human creators whose works formed part of training corpora and who have been economically displaced by AI-generated substitutes.

Recommendation 9 — AI Governance Authority: A dedicated AI Governance Authority -coordinating the Copyright Office, the DPDPA Data Protection Board, the Competition Commission of India, and sectoral regulators — should be established to provide coordinated regulatory oversight of AI-generated content across the creative, commercial, and public sectors.

Recommendation 10 — International Harmonisation: India should actively engage with WIPO’s AI and IP programme, the G20’s AI working group, and bilateral IP dialogues with the EU and US to shape an international consensus on AI copyright that reflects India’s interests as a major AI developer, user, and future rule-maker in global intellectual property governance.

XV. CONCLUSION

The Copyright Act, 1957, was enacted for a world in which only human beings could create original literary, artistic, and musical works. The anthropocentric authorship premise embedded in section 2(d), the human-centred “skill and judgment” standard endorsed by the Supreme Court in Eastern Book Company v D B Modak, and the entire conceptual architecture of copyright law — built upon the assumption of a creative human subject whose intellectual personality is expressed in the protected work — were adequate to the creative economy of the twentieth century. They are structurally inadequate to the generative AI era of the twenty-first century.

The challenge that AI-generated content poses to Indian copyright law is not marginal or speculative. It is present, commercial, and consequential today. AI systems are already generating advertising copy, legal memoranda, journalistic articles, musical compositions, and visual art that are commercially deployed across Indian industries. The absence of clear copyright ownership in such works creates legal uncertainty that impedes investment, generates contractual disputes, and disadvantages Indian AI developers and creative industries in global markets where competitors operate within more defined — if still evolving — legal frameworks.

This article has demonstrated through doctrinal analysis that section 2(d)(vi) does not, on a proper purposive construction, extend to autonomously AI-generated works. Parliament’s intention in 1994 was to attribute copyright in computer-generated works to the human who programmed the computer — not to vest rights in systems whose outputs are the product of autonomous machine learning processes over which no programmer exercises granular creative control. The Eastern Book Company “skill and judgment” standard confirms the doctrinal inadequacy: it presupposes a human author, and cannot be applied to AI-generated outputs without losing its analytical content.

The comparative survey reveals that no major jurisdiction has yet achieved a comprehensive solution. The UK’s CDPA 1988 s 9(3) provision, on which the Indian provision was modelled, is acknowledged by the UK IPO itself as inadequate. The US constitutional human authorship requirement forecloses the most obvious legislative solution. The EU AI Act imposes transparency obligations without resolving copyright ownership. China’s judicial approach is the most permissive but has not achieved legislative coherence. India is thus not following but competing with these jurisdictions — and has the opportunity to lead in designing a contextually calibrated, constitutionally grounded framework adapted to India’s constitutional values and development objectives.

The constitutional framework of Articles 14, 19, and 21 permits Parliamentary action to reform the copyright regime for AI-generated works. It does not require India to maintain the human authorship requirement as a constitutional absolute; it does require that any legislative solution be non-arbitrary, proportionate, and consistent with citizens’ rights to access information and engage in economic activity. The sui generis regime proposed in this article — with a reduced protection term, mandatory disclosure, a training data exception, and equitable remuneration for displaced human creators -satisfies these constitutional requirements while serving the commercial and innovation imperatives of India’s developing digital economy.

As India asserts its position as a global leader in artificial intelligence — through NITI Aayog’s National AI Strategy, the IndiaAI Mission, and the deployment of AI across governance, healthcare, and financial services — the inadequacy of its copyright framework for AI-generated content becomes not merely a legal lacuna but a strategic liability. World-class AI capability requires world-class IP infrastructure. Parliament’s enactment of a comprehensive, forward-looking framework for AI-generated intellectual property — incorporating the ten recommendations set out in this article — is both necessary and overdue. India’s emerging leadership in AI deployment makes it both appropriate and strategic to legislate the world-class framework that the generative AI era demands, positioning India as a thought leader in the global governance of artificial intelligence and intellectual property for the decades to come.

“The law must be responsive to the realities of the age it governs. A copyright framework premised upon the creative human author cannot govern a world in which machines generate the works of human culture at industrial scale. Legislative imagination is the only answer to

technological disruption.”

ENDNOTES

[1]   Copyright Act 1957 (Act 14 of 1957) (India) [hereinafter the Act].

[2]   Berne Convention for the Protection of Literary and Artistic Works (adopted 9 September 1886, entered into force 5 December 1887) 828 UNTS 221, art 2(1).

[3]   Copyright (Amendment) Act 1994 (Act 38 of 1994), s 3 (inserting s 2(d)(vi) into the Act).

[4]   Eastern Book Company v D B Modak (2008) 1 SCC 1 (India) [hereinafter Eastern Book Company].

[5]   Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS) (15 April 1994) 1869 UNTS 299, art 9 (requiring compliance with Berne Convention arts 1-21).

[6]   Berne Convention (n 2).

[7]   TRIPS (n 5).

[8]   WIPO Copyright Treaty (adopted 20 December 1996, entered into force 6 March 2002) 2186 UNTS 121.

[9]   Copyright, Designs and Patents Act 1988, s 9(3) (UK) [hereinafter CDPA 1988].

[10] Copyright Act 1957, s 2(d)(vi) (as amended by the Copyright (Amendment) Act 1994, s 3).

[11] CDPA 1988 (n 9), s 9(3).

[12] Copyright Act 1957, s 13(1)(a).

[13] Copyright Act 1957, s 17.

[14] Eastern Book Company (n 4) [34]-[38].

[15] CCH Canadian Ltd v Law Society of Upper Canada [2004] 1 SCR 339 (Canada).

[16] Navigators Logistics Ltd v Kashif Qureshi 2018 SCC OnLine Del 11303 (Delhi HC).

[17] R G Anand v Deluxe Films AIR 1978 SC 1613 (India).

[18] Indian Performing Rights Society Ltd v Eastern Indian Motion Pictures Association AIR 1977 SC 1443 (India).

[19] CDPA 1988 (n 9), s 9(3).

[20] Thaler v Perlmutter No 22-cv-01564 (DDC, 18 August 2023).

[21] US Copyright Office, “Copyright and Artificial Intelligence” Part 2: Copyrightability (2024).

[22] Feist Publications Inc v Rural Telephone Service Co 499 US 340 (1991) (establishing that copyright requires at leastminimal creativity by a human author).

[23] Regulation (EU) 2024/1689 of the European Parliament and of the Council on Artificial Intelligence (AI Act) [2024] OJ L1689/1.

[24] Beijing Internet Court, Tencent v Shanghai Yingxun Technology Co (2019) No 239 (China).

[25] Shreya Singhal v Union of India (2015) 5 SCC 1 (India).

[26] Constitution of India, art 19(1)(g).

[27] Internet and Mobile Association of India v Reserve Bank of India (2020) 10 SCC 274 (India).

[28] Maneka Gandhi v Union of India (1978) 1 SCC 248 (India).

[29] Justice K S Puttaswamy (Retd) v Union of India (2017) 10 SCC 1 (India).

[30] Ministry of Electronics and Information Technology, “National Strategy for Artificial Intelligence” (NITI Aayog,2018).

[31] Digital Personal Data Protection Act 2023 (Act 22 of 2023) (India).

Spoorthi B * BA LL.B. (Hons.) * University Law College, Bangalore University * 2026