Legal Recognition And Ownership Of AI- Generated Works Under Indian Copyright Law

Author: LawSphereX

Category: Intellectual Property

Published: 2026-08-23T20:31:32.539117+00:00

Updated: 2026-08-26T17:10:40.284+00:00

LEGAL RECOGNITION AND OWNERSHIP OF AI-

GENERATED WORKS UNDER INDIAN COPYRIGHT LAW

Priyadharshini G, LL.M., CHRIST (Deemed to be University), Bengaluru. Prof. Dr. Valarmathi R, CHRIST (Deemed to be University), Bengaluru.

ABSTRACT

The rapid advancement of generative artificial intelligence has fundamentally altered the landscape of creative production, posing unprecedented challenges to traditional copyright frameworks that are grounded in human authorship and individual creative expression. As generative systems increasingly produce literary, artistic, musical, and audiovisual outputs with minimal or indirect human intervention, existing legal doctrines struggle to accommodate questions of originality, authorship, ownership, and liability. This research critically examines how generative artificial intelligence disrupts foundational principles of copyright law and evaluates the adequacy of current legal responses to these emerging complexities.

The study explores the conceptual tension between humancentric originality standards and machine-generated creativity, highlighting the doctrinal uncertainty surrounding authorship attribution in AI-assisted and AI-generated works. It further analyzes competing claims of ownership among developers, deployers, and users of generative systems, demonstrating how traditional ownership rules fail to provide clarity in algorithm-driven creative processes. Particular attention is paid to copyright infringement risks arising from the use of protected works in AI training datasets and the potential for infringing outputs, raising complex questions of direct, secondary, and intermediary liability. In addition to legal uncertainty, the research addresses broader ethical and socio-legal concerns, including transparency, accountability, bias, and the erosion of human creative labour.

A comparative analysis of legal developments in jurisdictions such as India, the United States, the United Kingdom, and the European Union reveals fragmented and often inconsistent regulatory approaches to generative AI. Against this backdrop, the study argues for a balanced, human-centric regulatory framework that safeguards authors’ rights while fostering innovation. It proposes policy reforms emphasising transparency in training practices, calibrated liability standards, and statutory mechanisms to reconcile technological advancement with the normative objectives of copyright law.

Keywords: Copyright law, Generative Artificial Intelligence, Ownership, Copyright Infringement.

INTRODUCTION

The copyright law's basis has been challenged in ways that were never imagined before with the advent of advanced AI systems that can generate literary, musical, artistic, and audiovisual works. In light of the Indian Copyright Act, 1957, a work would only be protectable if it originates with a specific human author who has infused creativity and intellectual input into the expression of ideas. Yet, generative AI systems-large language models, image synthesizers, and music composers-can create commercially viable and aesthetically pleasing works with relatively little direct human involvement beyond supplying a prompt or configuring the AI models.

This new technological reality has shown that there is a significant gap in Indian copyright doctrine: when a work is created by an AI system, who, if anyone, possesses the exclusive ownership rights, and under what circumstances are such ownership rights enforceable. The question of AI-generated works is real and has important implications for the developers who seek capital investment, entrepreneurs whose business involves creating and operating AIassisted creative platforms, and for human creators who are concerned about the possible diminishing of their market and reputation interests. Indian copyright law says nothing on such questions today, neither statutorily recognizing AI systems as authors nor addressing what ownership should be between developers, owners of platforms, and end users who prompt and curate AI outputs.

In light of this, the silence of the law presents many stakeholders with a question of legal uncertainty, and Indian courts have not yet ruled definitively on whether machine-generated outputs are protected by the Act. The present chapter systematically considers the legal architecture under Indian copyright law governing ownership of works created with AI. The central contention is that Indian law requires reform in order to accommodate AI-assisted and AI-generated works through an explicit human-machine co-authorship framework, which roots authorship and ownership in discernable human contributions, without de facto or de jure legal personhood for AI systems.

It will draw distinctions between first, human-authored works whose creation was enhanced by AI tools, with copyright protection accruing in accordance with established norms; second, AI-assisted works where human creative direction is significant, with ownership resting in the

directing human actor(s); and third, works purely generated by AI, which, without legislative innovation or contractual allocation, remain unprotectable. By considering the wording of sections 17 - 21 of the Copyright Act, comparative foreign jurisprudence, and proposed reforms, this chapter plots a route to legal reform as part of a balance between innovation incentives and protection for human creative labor.

OWNERSHIP STRUCTURE UNDER THE INDIAN COPYRIGHT ACT, 1957

The first principle of copyright ownership on which the Indian Copyright Act, 1957 rests is that the author is the first owner of copyright in a work1. Section 17 of the Act provides that, subject to defined exceptions, the author of a work shall be the first owner of the copyright therein. 2The rule centers the threshold for questions of copyright ownership on authorship and provides an indispensable linkage between creative labour and property rights. 3The exceptions listed in section 17-for works made in the course of employment, commissioned photographs and cinematograph films, and Crown works-are themselves based upon conceptions of an author, since the exceptions merely shift ownership from one human agent to another (employer, commissioner or government)4.

Sections 18 to 21 address assignment and licensing of copyright; the law thus presupposes an existing owner who can transfer, license, or encumber rights5. The formal requirements imposed under these provisions - such as the requirement for assignment to be in writing, and to identify the work, territory, term, and nature of rights transferred - presuppose an author or legal entity capable of entering contractual arrangements and maintaining property interests. Significantly6, s 2(d) defines author in differential terms across categories of works: in relation to literary, dramatic and musical works, the author is the person who creates the work; for artistic works, the author includes the artist; for cinematograph films, the author shall be the producer; and for sound recordings, the author shall be the person who makes the arrangement to make the sound recording.

7Each of these definitions, however, presupposes a human agent

1 Indian Copyright Act, 1957, s.17.

2 Id.

3 See R.G. Anand v. Deluxe Films, AIR 1978 SC 1613 (establishing that copyright arises from authorship).

4 Indian Copyright Act, 1957, s.17 (enumerating exceptions that still presuppose authorship).

5 Id .s. 18-21.

6 See Who Owns AI-Generated Content? Copyright & Authorship Under Indian Law, Puthrans,

https://www.puthrans.com/who-owns-ai-generated-content-copyright-authorship-under-indian-law/

7 Indian Copyright Act, 1957, s. 2(d).

or identified legal entity, without express provision for non-human creators or automated systems. The addition of section 2(d)(vi) to the Copyright Act in 1994 states that in relation to any literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person who causes the work to be created. 8This provision, enacted decades before modern generative AI, was designed to address narrow categories of algorithmic output such as database compilations or routine data visualizations where the human role is limited to machine operation. 9The phrase causes the work to be created has been interpreted by Indian legal scholars as referring to the person who finances, directs, or exercises overall control over the computational process-usually an employer, company or contractor-rather than the enduser who merely inputs prompts or parameters.

10Yet section 2(d)(vi) provides minimal guidance for complex scenarios where multiple human actors-developer, trainer, operator, userexercise varying degrees of influence over an AI-generated output and where the system's autonomous processing capacity makes it unclear who, if anyone, causes the final work to emerge.

THE ORIGINALITY REQUIREMENT AND ITS APPLICATION TO AI OUTPUTS

A work gets copyright protection under section 13 of the Copyright Act only when it falls

within the enumerated categories as a literary, dramatic, musical, or artistic work and also meets the minimum standard of originality. Originality, referring to independent creation and a modicum of human intellectual effort, skill, or judgment, has been defined by Indian copyright jurisprudence following international precedent, not as novelty in ideas. The Supreme Court laid down in R.G. Anand v. Delux Films that there is copyright in an idea as expressed but not in an idea simpliciter, and that infringement must be based upon substantial similarity in mode of expression rather than mere coincidence or sameness of theme11. From this follows the conclusion that originality attaches to human choices choice of subject matter, determination of form, exercise of editorial judgment rather than to the complexity or impressiveness of a work.

8 Id.s. 2(d)(vi).

9 See Navigating Indian Copyright Framework in the Age of AI generated Works, CSIPR, NLIU (discussing section 2(d)(vi) and its limitations).

10 See AI and Digital Age: Commencement of New Era of Copyright Infringements, Naya Legal,

https://www.nayalegal.com/ai-and-digital-age-commencement-of-new-era-of-copyright-infringements.

11 See R.G. Anand v. Deluxe Films, AIR 1978 SC 1613 (establishing minimal originality threshold).

When applied to AI-generated works, the originality requirement becomes quite problematic. Where an AI system generates a novel, poem, or image with only limited and general human input, say, in the form of selecting a model and pushing generate, it is hard not to notice that the question of whether originality can be attributed to the resulting work at all, or whether any originality resides, if anywhere, in the model's design and training rather than in the immediate generative act. 12The Delhi High Court's decision in Navigators Logistics Ltd. v. Kashif Qureshi & Ors. held that a compilation of client data and information lacked copyright protection because it was merely a product of labor and mechanical arrangement, devoid of the requisite skill or creative judgment necessary to meet the originality threshold.

13By analogy, a completely machine-generated output reflecting no human author's intellectual creation might thus fail the originality bar, and such works could fall de facto into the public domain despite commercial or aesthetic value. Currently, Indian courts haven't fully addressed the matter in a reported judgment that involved generative AI, yet the existing legal doctrines have already made their implications apparent: copyright protection for most completely autonomous AI outputs would be ruled out by a strict originality requirement if applied with the utmost severity.

This situation would lead to a protection deficit whereby a big portion of economically and culturally important AI-generated content such as advertising campaigns, software codes, and artistic works, would not be able to rely on copyright protection at all, hence, there would be no legal outlet against unauthorized reproduction or adaptation and no incentive for investment in AI development. Conversely, a more lenient originality criterion that grants protection to robots' outputs based solely on their functional usefulness or aesthetic appeal might cut copyright off from its underlying humancreativity justification and India might find itself granting monopolies over the whole class of content produced by machines that would otherwise be better off unclaimed and thus, in the public domain.

JUDICIAL INTERPRETATION OF AUTHORSHIP AND ITS LIMITATIONS IN THE

AI CONTEXT

Even though Indian courts have created an extensive body of case law on the subjects of authorship, originality, and copyright ownership, their rulings did not take into account AI-

12 Artificial Intelligence And Copyright Law in India, supra note 5.

13 Navigators Logistics Ltd. V. Kashif Qureshi & Ors., (2018) 254 DLT 307 (Delhi High Court) (holding that mere mechanical compilation lacks copyright protection).

generated works and were based most of the time on the human agency aspects that come up as issues when dealing with autonomous systems. 14 In the case of Amar Nath Sehgal v. Union of India, the Delhi High Court stated that the moral rights of the author are closely connected to the author’s personality, dignity, and creative work, thus calling them the soul of the author's works. 15This line of reasoning implicitly suggests that the author of a work is a human who has a reputation, honor, and personal attachment to his or her creative outputassumptions that do not apply to AI systems at all. In decisions relating to copyright in software and compilations, courts have also emphasized the role of a human intellectual effort to transform raw inputs into a protectable expression.

16Where courts have granted protection, they have invariably found a human author whose skill, judgment, and creative choices shaped the end product. By contrast, where courts have denied protection such as in Navigators Logistics they have done so on the basis that the defendant failed to establish the requisite human author and exercise of skill and judgment. 17 The text of the law itself is thus the primary restricting factor in judicial evolution. The Copyright Act's Sections 2(d) and 17 establish the concept of authorship in purely human terms and, although the courts have a bit of a leeway to interpret it, they cannot change the basic statutory concepts without the legislature's permission.

18 A court might extend the meaning of authorship to include the arranger for an AI system to be treated as an author by referring to section 9(3) of the United Kingdom's Copyright, Designs and Patents Act 1988, but this would require much reasoning to the effect that it would not be in conflict with the law's framework and the ruling in the R. G. Anand case. 19Without a clear indication from the legislature, the Indian judiciary is left with a dilemma: either deny protection to works of AI origin and at the same time risking the opening of a legal vacuum for the downstream output that has

14 See Legal Implications of AI-Generated Works in Copyright Law: An Analysis of Raghav Artificial

Intelligence v. Union of India, Lawful Legal, https://lawfullegal.in/legal-implications-of-ai-generated-works-incopyright-law-an-analysis-of-raghav-artificial-intelligence-v-union-of-india-2024/ 15 Amar Nath Sehgal v. Union of India, (2005) 30 PTC 228 (Delhi High Court) (characterizing moral rights as the soul of authorship).

16 See R.G. Anand v. Deluxe Films, AIR 1978 SC 1613.

17 Navigators Logistics Ltd. V. Kashif Qureshi & Ors., (2018) 254 DLT 307.

18 See Case Analysis: Amar Nath Sehgal v. Union of India (2005), LegalBites,

https://www.legalbites.in/category-intellectual-property-rights/case-analysis-amar-nath-sehgal-v-union-of-india- 2005-moral-rights

19 See The UK’s Curious Case of Copyright for AI-Generated Works, Authors Alliance,

https://www.authorsalliance.org/2025/05/19/the-uks-curious-case-of-copyright-for-ai-generated-works-whatsection-93-means-today/

commercial value or adopt temporary measures that will make technology developers, investors, and creative professionals uncertain and inconsistent in their practice.

THE PREREQUISITES OF LEGAL PERSONHOOD IN COPYRIGHT DOCTRINE

Copyright doctrine has long proceeded on the premise that authors are either natural personsthat is, human beings-or legal persons-that is, corporations, governments, institutions-capable of maintaining property, entering into contracts, bearing legal obligations, and wielding agency within the system of law. 20This rule of legal personhood serves a number of functions: it enables the identification of a party that can enforce rights through litigation and enter licensing and assignment agreements and bear responsibility for any duties or liabilities associated with ownership of a copyright.

21To confer copyright upon an entity that lacks such legal personhood would involve a conceptual and practical contradiction: a right-holder invisible to the law-one that cannot sue infringers, cannot license others, cannot defend against counterclaims, and cannot be held accountable for infringing conduct by others. Other jurisdictions have also seen proposals for granting legal personhood to AI systems, most famously the possibility of limited electronic personhood for autonomous robots considered by the European Union. 22 But even in those discussions, the motivation for personhood is not that AI systems bear any resemblance to conscious, intentional moral agents but rather to make it possible to identify a responsible agent for the purposes of assigning liability or property rights.

23Today's large language models and image generators - let alone any other current AI systems - do not have subjective experience, self-awareness, or even autonomous interests justifying considering them moral agents and as such entitled to rights. 24 They are, in the current state of technology, tools albeit powerful and semi-autonomous tools that operate under control of human beings, even when that control is distributed, probabilistic, or executed after initial training.

20 See Monkeying with Copyrights: Who Owns the Monkey Selfies, Jaburg Wilk,

https://www.jaburgwilk.com/news-publications/monkeying-with-copyrights-who-owns-the-monkey-selfies-alesson-in-copyright-ownersh

21 Id.

22 See Generative AI and Copyright: Training, Creation, Regulation (European Parliament Policy Dept. 2025). 23 See AI Act und Copyright, Taylor Wessing, https://www.taylorwessing.com/en/insights-andevents/insights/2024/05/ai-act-und-copyright

24 See Non-Human Authorship, in AI, Copyright and Data Privacy in Education (Pressbooks 2024).

NARUTO V. SLATER MONKEY SELFIE CASE

The 2018 decision of the United States Court of Appeals for the Ninth Circuit in Naruto v. Slater provides cautionary precedent for any jurisdiction considering extension of authorship or copyright rights to non-human entities. 25 In that case, a macaque monkey named Naruto accidentally took a series of selfie photographs with wildlife photographer David Slater's camera on a field trip to Indonesia. The animal rights organization People for the Ethical Treatment of Animals (PETA) filed suit on Naruto's behalf as his next friend, arguing that the monkey, as the being who pressed the camera shutter and exercised creative control over pose and framing, should be considered the author of the photographs and copyright owner.

The Ninth Circuit affirmed the district court's dismissal, holding that the Copyright Act, by its express terms, grants statutory standing to sue only to natural persons and legal entities recognized by law, and that an animal-lacking legal personhood-cannot be vested with copyright or authorized to bring an infringement action. 26 The court emphasized that while PETA's arguments about Naruto's factual role in creating the images possessed some superficial appeal, copyright law does not turn on who causally produced a work but on who holds legal personhood and the authority to exercise property rights. 27 The court further noted that extending copyright to non-human actors would require clear congressional action rather than judicial gap-filling, and that policy arguments about fairness or efficiency could not override statutory design.

The Naruto case has become an anchor point in debates about AI authorship, with courts and commentators finding that the same reasoning that a non-human entity lacks statutory standing and that legal personhood is a condition for copyright ownership applies with equal or greater force to AI systems. While Naruto was a case where animal action got control of a camera at an unintended moment, A. I. works under programming and parameters set by humans; while Naruto might have somewhat materially benefited from copyright protection and been damaged by infringement, A. I. cannot; and while it had no line of human causation it could attribute (without relying on the personhood of the animal), there is a clear line license of human causation for attributing copyright to the developer, trainer or A. I. tutor.

Indian courts, operating under a similarly human-centered statutory framework, will be likely to use any of

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

26 Id. At 41-13.

27 Naruto v. Slater, supra note 25

this reasoning and refuse to recognize A.I. systems as authors without an affirmative statutory adoption28.

THE INDIAN CONTEXT: THE RAGHAV CASE AND POLICY DISCUSSIONS

Although India has not issued a binding judicial precedent recognizing AI as an author, publicized efforts to register AI-generated works have spurred debate and policy attention. 29 The most prominent example involves Ankit Sahni, an intellectual property lawyer who developed an AI-based artistic tool called RAGHAV (Robust Artificially Intelligent Graphics and Art Visualizer) and sought to register artwork created with RAGHAV's assistance at the Indian Copyright Office. 30Sahni's first application listed RAGHAV as the sole author, but the Copyright Office rejected it, citing the requirement under section 2(d) of the Copyright Act that an author must be a person or entity capable of bearing copyright ownership.

A subsequent application, filed in 2020, listed both Sahni and RAGHAV as co-authors; this application was granted registration, becoming a widely citedthough contestedexample of AI co-authorship recognition in India. The RAGHAV registration has been called into question by legal scholars and the U. S. Copyright Office, which denied Sahni's application for registration of a similar work in the United States for insufficient human authorship under U. S. law. Since the decision, Indian Copyright Office guidance has emphasized that even though AI can be a tool in the creative process, it is only the human or other legal entity that causes the work to be created under section 2(d)(vi) that can be vested with authorship and copyright ownership, and not the AI system itself.

The registration of RAGHAV as co-author thus appears to reflect administrative practice rather than settled legal doctrine, and the status remains uncertain pending litigation or legislative clarification. In 2025, the Government of India, under the aegis of the Ministry of Commerce, constituted an expert committee to conduct a review of copyright law in the light of AI developments. 31 This committee is accordingly obliged to consider, amongst others, definitional uncertainties

28 See Monkeying with Copyrights: Who Owns the Monkey Selfies, supra note 26

29 See Exclusive: India recognises AI as co-author of copyrighted artwork, Managing

IP.https://www.managingip.com/article/2a5bqo2drurt0bxl7ab24/exclusive-india-recognises-ai-as-co-author-ofcopyrighted-artwork.

30 Id.

31 See Government Forms Expert Panel to Review Copyright Law Amid AI Disputes, Varindia,

https://www.varindia.com/news/government-forms-expert-panel-to-review-copyright-law-amid-ai-disputes.

with respect to the authorship of AI-generated works, assessment of ownership rights between the developers and the user, applicability of fair use and licensing schemes to both the AI training as well as outputs generated, and assimilation of Indian law with the international regime.32These undertakings of the committee constitute an acknowledgment at the policy level that AI-generated content falls outside the scope of the current Copyright Act; however, such reform remains pending in the legislature. Unless such reform materializes, the legal status of AI authorship remains unclear in India, which again creates uncertainty for technology companies, creative professionals, and investors who operate at the intersection of AI and copyright33.

CONCEPTUAL BASES FOR DEVELOPER AND USER OWNERSHIP CLAIMS

When a generative AI system produces an original work-be it a text, image, musical composition, or hybrid output-several human actors can articulate plausible claims to authorship and ownership based on their respective contributions to the generative process. 34 The AI developer or company that designed, trained, and deployed the model might claim that it is the owner of copyright in outputs because the model's architecture, training data curation, and inference algorithms represent the primary creative and technical labor underpinning any output the model could produce. 35Viewed this way, individual users are merely operators of an ingenious tool, akin to photographers wielding a camera or composers playing a synthesizer, and the user's contribution by way of prompting, parameter selection, or curation cannot suffice to displace ownership from the developer.

On the other hand, the end user who crafts detailed prompts, iteratively refines outputs, and makes editorial judgments about which generated variants to retain may make a strong case for being the true author of the final work. 36The users argue that the AI model is nothing but an advanced tool of human expression, their prompts encode unique creative intent and curatorial 32 See India Sets Up Expert Panel to Review AI-Era Copyright Laws, Licit360, https://licit360. in/india-formsexpert-panel-to-review-copyright-laws-in-the-ai-era/ 33 See Ankit Sahni’s AI Co-authored Artwork Denied Registration by US, Continues to be Registered in India, Spicy IP, https://spicyip. com/2023/12/ankit-sahnis-ai-co-authored-artwork-denied-registration-by-us-continuesto-be-registered-in-india. html

34 See Navigating Ownership Over AI-Generated Work, The IP Press,

https://www.theippress.com/2025/06/16/artificial-intelligence-and-copyright-navigating-ownership-over-aigenerated-work/ 35 See Artificial Intelligence and copyright: use of generative AI tools to..., European Commission Intellectual Property Helpdesk, https://intellectual-property-helpdesk.ec.europa.eu/

36 See Who Owns AI-Generated Content?, supra note 11.

judgment, and such outputs would either not exist or be radically different without their guiding vision.37 In this framework, the developer is akin to a tool manufacturer, and while a tool manufacturer might retain intellectual property in the tool itself, they certainly don't own the works created using the tool by independent users. Indian copyright law, written well before the rise of generative AI, offers no direct guidance for reconciling this clash of developer and user claims.38 Neither the person who causes the work to be created language of section 2(d)(vi) nor the broader authorship provisions of section 2(d) clearly assigns ownership between entities that jointly contribute to a work's production in fundamentally different ways-one through technical architecture and training, the other through expressive prompting and curation39.

CONTRACTUAL ALLOCATION OF OWNERSHIP RIGHTS VIA PLATFORM

TERMS OF SERVICE

In real situations, between developers and users, the matter of ownership is settled through the contractual terms of service and license agreements that regulate user access to and use of AI platforms. Several top generative AI platforms-including those supporting image creation, text writing, and code synthesis-have clauses that determine who will own the copyright of the materials generated by AI. They usually allow users to have very wide rights to export commercially the outputs while keeping the developers' ownership of the underlying model, training data, and platform infrastructure.

These contracts are made in such a way that the developers have a practical view that the users are going to expect giving them ownership or at least rights to commercialize the outputs they prompt and that such rights are going to attract more people to the platform and ultimately make them happy users40. Indian courts have traditionally given effect to intellectual property clauses in digital services agreements, provided they do not violate mandatory statutory provisions or public policy. 41 Courts acknowledge that parties can contractually allocate copyright ownership in ways that differ from statutory defaults, so long as the allocation respects the requirement that copyright 37 See AI Generated Artworks & Copyright, Surana & Surana, https://suranaandsurana. com/ai-generatedartworks-copyright/ 38Artificial Intelligence And Copyright Law in India, supra note 5.

39 See Ownership of AI generated content, supra note 4.

40 See Artificial Intelligence and copyright: use of generative AI tools to..., supra note 50.

41 See Section 57 of Copyright Act: An Overview of Author’s Moral Rights, The Legal School,

https://thelegalschool.in/blog/section-57-of-copyright-act

vest in a human author or legal entity. 42However, contractual allocation cannot create copyright where none exists under statute. If an AI-generated output fails to meet the originality threshold - or lacks identifiable human authorship - no contractual term can retrofit statutory copyright protection onto an unprotectable output; the contract can only allocate ownership of such benefits as do exist, or create private rights to prevent downstream copying without invoking copyright law. 43 In the context of generative AI, contractual clarity regarding ownership allocation might reduce uncertainty between developers and users and may signal industry norms which legislators subsequently consider codifying into statutory form.

44However, such contractual frameworks remain fragile in the absence of statutory guidance given that courts may question whether the parties truly intended to allocate copyright in works which may not qualify for statutory protection, or interpret ambiguous allocation clauses conservatively in favour of the statutory default - developer as author45.

ANDERSEN V. STABILITY AI

Litigation and Implications for Indian Law The high-profile US litigation of Andersen v. Stability AI Ltd. , in which visual artists filed a group action against generative AI companies for copyright infringement, casts light on the legal battle between developers and users while raising key questions with respect to the responsibility of developers in model training and its subsequent use46. In October of 2023, the Court in the Northern District of California found that visual artists Sarah Andersen, Kelly McKernan, and Karla Ortiz had sufficiently alleged direct and vicarious copyright infringement claims against Stability AI, a developer of the Stable Diffusion image generator, by non-consensually incorporating billions of copyrighted images into the model's training dataset.

The court found sufficient evidence showing that Stability AI had accessed and compressed the artists' works into the model's parameters, allowing the system to eventually generate artworks mimicking the artists' distinctive styles.

42 See Navigators Logistics Ltd. V. Kashif Qureshi & Ors., (2018) 254 DLT 307.

43 Who Owns AI-Generated Content?, supra note 11.

44 See Andersen v. Stability AI, Knowing Machines, https://knowingmachines.org/knowing-legalmachines/legal-explainer/cases/andersen-v-stability-ai 45 See Copyright in the Age of Generative AI, Part II, Culpa Law Review, https://www.culawreview.org/currentevents-2/copyright-in-the-age-of-generative-ai-part-ii-reinterpreting-dmca-1202-and-encoded 46 See Andersen v. Stability AI: The Landmark Case Unpacking the Copyright Risks of AI Image Generators, JIPEL, https://jipel.law.nyu.edu/andersen-v-stability-ai-the-landmark-case-unpacking-the-copyright-risks-of-aiimage-generators/

While Andersen's complaint primarily deals with infringement of the training data, rather than ownership of outputs, the case illustrates several principles relevant to ownership of AI outputs. First, the case suggests developers may be held directly or secondarily liable based on how their models are trained and used-meaning their responsibility does not stop at passively providing a tool, but instead requires vigilance regarding unauthorized incorporation of content. Second, the case illustrates that outputs from generative models may themselves be infringing derivative works where they closely reproduce pre-existing copyrighted material, and thus create exposure both for the developer and any users of the model.

47Third, litigation has spurred a closer look at user behavior in prompting an AI system to create works in the style of particular artists, raising the question of whether the user may be held liable for inducing infringement via their prompting behavior.. 48Instead, Andersen recommends that Indian law focus efforts on the chain of responsibility among developers, trainers, and users rather than seeking to identify a metaphysical author by attempting to locate creative agency.

In such a scheme, developers would primarily be responsible for making sure training data does not infringe copyright and that the model is designed in a way so as not to violate the intellectual property rights of third parties; users would be responsible for making sure their prompts do not ask the system to generate infringing outputs; and courts would ascribe ownership over non-infringing outputs based on who exercised meaningful creative control and expressive intent. Such an approach would cohere with the principles underlying Indian copyright, emphasizing human authorship while acknowledging the distributed agency inherent in AI-assisted production.

NATURE, SCOPE AND RATIONALE OF MORAL RIGHTS UNDER INDIAN LAW

SECTION 57

Copyright Act gives creators Author's Special Rights, which includes two moral rights: (i) the right of paternity, which is the right of authors to claim authorship and to have their name associated with the work; and (ii) the right of integrity, which is the right of authors to object to distortion, mutilation, modification, or derogatory treatment prejudicial to their honor or reputation.49 In important respects, these are inalienable rights, surviving the assignment or

47 Id.

48 See Copyright in the Age of Generative AI, Part II, Culpa Law Review, https://www.culawreview.org/currentevents-2/copyright-in-the-age-of-generative-ai-part-ii-reinterpreting-dmca-1202-and-encoded

49 See Section 57 of Copyright Act: An Overview of Author’s Moral Rights, The Legal School,

https://thelegalschool.in/blog/section-57-of-copyright-act

licensing of economic copyright and continuing beyond the term of the work into the public domain. 50The Indian courts have construed Section 57 expansively and explained that moral rights are not confined to protection of purely economic interests but extend to the protection of an author's dignity, personality, and creative identity. 51 The landmark judgment in Amar Nath Sehgal v. Union of India embodies the personality-based foundation of Indian moralrights doctrine. 52There, the Delhi High Court ruled that the government's decision to remove, store, and alter a monumental sculpture by the famous artist Amar Nath Sehgaldespite Sehgal's economic copyright being transferred to the governmentinfringed Sehgal's moral rights under section 57.

53The court described moral rights as the soul of the author's works, declaring that an author has a right to preserve, protect and nurture his creations through his moral rights and that a creative individual is uniquely invested with the power and mystique of original genius, creating a privileged relationship between a creative author and his work. 54 The latter reasoning grounds moral rights in the author's personal connection to, investment in, and reputation regarding the work connection that assumes human subjectivity as well as susceptibility to injury by way of misuse or distortion.

CONCEPTUAL INCOMPATIBILITY OF MORAL RIGHTS WITH NON-HUMAN

AUTHORS

This personality and dignity-centric rationale behind moral rights therefore creates an insurmountable tension when extended to AI-created works: because the AI systems lack consciousness, subjective experience, reputation, or feelings that could be harmed through modification or misuse, they cannot meaningfully exercise moral rights or suffer prejudice to honor through derogatory treatment. 55An AI system cannot feel offended by distortion of its outputs, cannot claim that its reputation has been damaged, and cannot assert that its creative vision has been betrayed-all essential components of integrity-rights claims as understood in Indian law. 56Further, the very notion of an AI system claiming authorship or wanting attribution, which is what the paternity right gives the creator, is incoherent when extended to 50 Indian Copyright Act, 1957, s.

57; see also Moral Rights under Copyright Laws: A Peep into Policy, Spicy IP, https://spicyip. com/2007/12/moral-rights-under-copyright-laws-peep-2. html

51 See Amar Nath Sehgal v. Union of India, (2005) 30 PTC 228.

52 Id.

53 Id.; see also Amar Nath Sehgal v. Union of India: Aiding the Delhi High Court in Recognizing Moral Rights in India, Anand & Anand, https://www.anandandanand.com/pdfgen/4968/

54 Amar Nath Sehgal v. Union of India, (2005) 30 PTC 228.

55 See Moral Rights in Media and Entertainment Law, IP and Legal Filings,

https://www.ipandlegalfilings.com/moral-rights-in-media-and-entertainment-law-a-contemporary-analysis/

56 Id.

a non-human creator. 57Attribution exists both to ensure that human creators receive credit, recognition, and reputational gain from their work, and to provide audiences with context on the basis of which they can make informed decisions about authorship and influence. For AI systems, attribution performs no analogous role: it is the public who are served by transparency in terms of AI involvement in creation, and the AI system has no moral or reputational interest in attribution. 58 Recent Indian scholarship on AI-generated artworks stresses that moral rights under section 57 are based on human authorship, and any attempt to afford moral rights to the AI systems would reduce those rights to an empty legal formality or else force an incoherent anthropomorphization of algorithms.

59There are some commentators who propose that the approach must be directed towards recognizing moral rights of human contributors to the AIassisted works-for example, the artists who substantially edit or incorporate the AI outputs, data curators whose choices affect the composition of the training data, or users who exercise substantial creative control-even in that process when parts of the work are created by automated processes. This methodology retains the human-centered justification for moral rights and extends this concept to enable AI contributions to stand along with current creativity.

PROSPECTIVE JUDICIAL APPROACHES

Moral Rights in AI-generated Works Should Indian courts face moral-rights claims of AIgenerated or AI-assisted works, a number of interpretive routes are open that preserve the human-centered cast of Section 57 while responding to technological change. 60Courts may acknowledge moral rights of identifiable human contributors-developers, trainers, editors, or users-whose creative input and persona are invested in AI-assisted works and whose reputation may be affected by how the work is subsequently used or modified.

61For example, an artist who has adopted an AI tool to create a rough image and then made substantial modifications to that image and incorporated and transformed the image into a finished piece of artwork may still claim moral rights in the final work, even if AI-generated at intermediate stages, based on the fact that her final creative choices and identity are present in the final expression62. On the other hand, moral-rights jurisprudence may extend protection to human creators whose unique

57 See Section 57 of Copyright Act: An Overview, supra note 49.

58 See Copyright Protection for AI-Generated Works in India.

59 Legal Implications of AI-Generated Works in Copyright Law, supra note 14

60 Moral Rights in Media and Entertainment Law, supra note 55.

61 Id.

62 See Case Analysis: Amar Nath Sehgal v. Union of India (2005), LegalBites,

https://www.legalbites.in/category-intellectual-property-rights/case-analysis-amar-nath-sehgal-v-union-of-india- 2005-moral-rights.

styles, works, or creation processes were part of the AI training dataset and were reproduced or closely reproduced in an AI output without permission, treating them as unconsented alterations of the original creator's rights. 84 Under this approach, if an AI model is trained on a visual artist's copyrighted works and then produces images closely mimicking that artist's distinctive style, the original artist could argue that the produced images are distorted, unauthorized derivatives that prejudice her reputation by implying endorsement or authorship of works she did not create or approve. 63 This theory would require courts to extend integrity rights beyond traditional infringement and into a kind of personality right against unauthorized mimicry, a development that would be novel but arguably consistent with the protective impulse underlying section 57.

Another approach would be for courts to embrace a safe harbor for works solely the result of automation; a court so ruling could conclude that works wholly unconnected with a human contributor-that is, which could have no identifiable human authorstand entirely outside the scope of section 57, since the rights conferred by section 57 implicitly assume a human author. 64This approach would maintain section 57 as conceptually coherent while leaving open space for legislative innovation on rights in AI-generated works. Within this logic, when and if Parliament acts to create new protectable categories of AI-generated works or AI systems as rights holders, the courts would revisit the application of section 57; to that point, exclusively machine-generated works fall outside the moral-rights regime.

Hawking also had deep insight into the role that gravity plays in shaping the internal structure of black holes, and so predicted that the entropy for a black hole is proportional to its surface area rather than its volume.

COMPARATIVE INSIGHTS: UK, US, AND EU APPROACHES TO OWNERSHIP

THE UNITED KINGDOM

Section 9(3) of the CDPA 1988 and the Arrangements Necessary Standard The United Kingdom has long had the most explicit statutory treatment of computer-generated works among common-law jurisdictions, providing a potential template though also a cautionary tale for Indian reform efforts.65Section 9(3) of the Copyright, Designs and Patents Act 1988 designates the person by whom the arrangements necessary for the creation of the work are 63 Andersen v. Stability AI: The Landmark Case Unpacking the Copyright Risks of AI Image Generators, JIPEL. https://jipel.law.nyu.edu/andersen-v-stability-ai-the-landmark-case-unpacking-the-copyright-risks-of-ai-imagegenerators/.

64 Legal Implications of AI-Generated Works in Copyright Law, supra note 14.

65 The UK’s Curious Case of Copyright for AI-Generated Works, supra note 19.

undertaken as the author of a computer-generated literary, dramatic, musical, or artistic work. 66This provision, enacted in 1988 well before modern generative AI, was designed to address algorithmic outputs where no human author could be identified but where copyright protection seemed desirable for policy reasons, and it represented a deliberate departure from the traditional human-authorship requirement. 67 Under UK law, computer-generated work is defined in section 178 as work generated by computer in circumstances such that there is no human author of the work.

68 Once a work qualifies as computer-generated under this definition, copyright duration is reduced to 50 years from creation (rather than the standard 70 years postmortem auctoris), and moral rights are not available to the imputed author, reflecting the law's recognition that a non-human entity cannot meaningfully exercise such rights. 69The arrangements necessary language has been interpreted in the sole reported case addressing section 9(3) before 2020, Nova Productions Ltd. v. Mazooma Games Ltd. , where the High Court held that a video game programmer who designed the appearance of composite frames and the rules governing their generation was the person making the necessary arrangements and thus the author.

70 Although section 9(3) has been invoked as a possible model for Indian reform-and indeed some expert panel discussions have referenced the provision-it has attracted significant scholarly criticism and is now the subject of governmental review in the UK itself. 71Critics observe that section 9(3) either lacks clear normative justification or else operates incoherently: if originality (as it is usually understood) requires human authorship, then labeling a nonhuman output as authored by someone who made arrangements for its creation seems to define authorship away from originality; otherwise, if originality is not required for works under section 9(3), then the provision risks giving copyright monopolies to routine, unoriginal machine output that might better serve the public interest if left in the public domain.

72What's more, identifying the person making arrangements becomes increasingly difficult with sophisticated generative AI, where responsibility may be distributed across model developers, training-data curators, platform operators, and end-users who supply prompts and

66 The UK’s Provisions on Computer-Generated Works, European Copyright Society (2018).

67 The UK’s Curious Case of Copyright for AI-Generated Works, supra note 19.

68 The UK’s Provisions on Computer-Generated Works, supra note 66.

69 See Ownership of AI-generated content in the UK, A&O Shearman,

https://www.aoshearman.com/en/insights/ownership-of-ai-generated-content-in-the-uk (Aug. 19, 2024).

70 The UK’s Provisions on Computer-Generated Works, supra note 66.

71 See The Curious Case of Computer-Generated Works (CGW) in THJ Systems v. Sheridan, City Law Forum, https://blogs.city.ac.uk/citylawforum/2024/04/17/the-curious-case-of-computer-generated-works-cgw-in-thjsystems-v-sheridan/ (Apr. 16, 2024).

72 See The UK’s Curious Case of Copyright for AI-Generated Works, supra note 209.

parameters.73

THE UNITED STATES

Human-Authorship Requirements and Copyright Office Guidance The United States has adopted a markedly different approach, relying on strict human authorship requirements and withholding copyright protection from works generated completely by AI without meaningful human creative input. 74The U. S. Copyright Office has issued several policy statements and registration decisions confirming that human authorship is a bedrock principle of U. S. copyright law and that works created solely by a machine or product of nature are not eligible for copyright protection. 75In March 2023, the Copyright Office issued guidance that confirmed its position that if a work contains more than a de minimis amount of AI-generated material, applicants are required to disclose that fact and describe the human author's specific creative contributions.

That position was reinforced in January 2025 when the Copyright Office released its comprehensive report on Copyright and Artificial Intelligence, reiterating that human authorship remains the threshold requirement for copyrightability and that purely AIgenerated material created through prompting alone does not qualify for protection. The Office further clarified that, while relying on AI tools to enhance human creativity-such as using AI to suggest ideas or edit images prior to final human-authored work-does not bar copyright, the mere provision of detailed prompts without substantial human creation of the resulting output is insufficient to generate copyright. That position has been upheld in litigation, most prominently in Thaler v. Perlmutter, where the U. S.

District Court for the District of Columbia affirmed the Copyright Office's refusal to register an image created entirely by an AI algorithm without human authorship. The U. S. approach retains a clear bright-line rule-human authorship is required-while allowing incremental case-by-case analysis of what constitutes sufficient human contribution to qualify for protection in works incorporating AI-generated material. 76This approach avoids the conceptual incoherence of section 9(3) by preserving the view that copyright is based on human creativity, not simply on human involvement or direction. However, it also means that many commercially valuable AI-generated works fall 73 See Inside the Copyright Office’s Report, Copyright and Artificial Intelligence, Part 2: Copyrightability, Library of Congress, https://blogs. loc.

gov/copyright/2025/02/inside-the-copyright-offices-report-copyright-andartificial-intelligence-part-2-copyright/.

74 See Copyright Office Publishes Report on Copyrightability of AI-Generated Works, Skadden,

https://www.skadden.com/insights/publications/2025/02/copyright-office-publishes-report.

75 See U.S. Copyright Office Issues Report on Artificial Intelligence and Copyrightability,

https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.

76 Id.

outside of copyright protection, with developers instead relying on trade-secret protection, contractual constraints, or suigeneris database rights to safeguard their investments.77

THE EUROPEAN UNION

AI Act, DSM Directive, and Emerging Regulatory Frameworks The European Union has not established sui generis authorship rules for AI-generated works but instead relies on existing originality standards rooted in the personal intellectual creation of an author, while layering atop copyright law a comprehensive regulatory framework for AI systems. EU copyright doctrine, molded by CJEU precedent, requires that protected works reflect the author's own intellectual creation and flow from their free and creative choices, a standard that presumes human authorship and excludes purely automated outputs from protection. At the same time, the EU AI Act, which took effect on August 1, 2024 and will become largely applicable by August 2026, imposes new transparency and accountability obligations on developers of general-purpose AI models such as ChatGPT and Stable Diffusion.

Recital 106 of the EU AI Act insists that providers of GPAI models will have to respect EU copyright, including the exceptions for text and data mining introduced by the Digital Single Market Directive (Directive 790/2019/EU). The DSM Directive offers two binding copyright exceptions for TDM, one in favour of research organisations and cultural heritage institutions (Article 3), and a second one for any beneficiary (Article 4), with an opt-out right for holders of copyright and/or related rights.

It goes without saying that these TDM exceptions are narrow, do not explicitly allow the use of copyright protected works for the training of datasets in view of developing commercial AI models, but rather anticipate the uses of lawfully accessed content for the purposes of computational analysis aimed at the extraction of patterns and trends, with right holders being allowed to opt out via machine-readable means. 78In reality, the opt-out mechanism of the DSM Directive has been difficult to implement in practice, due in large part to the absence of a default standard for machine-readable opt-outs and the general inability of copyright holders to signal their preferences to developers of AI tools.

While EU policymakers have signaled a willingness to eventually establish clearer rules and possibly a licensing market for training data, thus allowing for copyright compliant development of AI, the situation remains unsettled as of late 2024. A study from 2024-25 for the European Parliament finds that 77 U. S. Copyright Office Issues Report on Artificial Intelligence and Copyrightability, supra note 75 78 See Summary: The Text and Data Mining Exception in Copyright and Related Rights Digital Single Market, 4IP Council, https://www. 4ipcouncil. com/research/summary-text-and-data-mining-exception-under-copyrightand-related-rights-digital-single-market .

EU Member States have a common understanding that works solely created by AI should not be copyright-protected, as international copyright treaties grant the status of authors only to natural persons - under the Berne Convention. The EU approach therefore diverges from both the UK and US models in that it neither extends copyright to non-human entities nor needs special statutory fiction, such as s 9(3), in order to parcel out authorship, nor does it seek to issue general guidance regarding copyrightability of works generated with the aid of AI (as has the US Copyright Office). Instead, it keeps the same standard regarding originality, while adding more regulatory requirements in respect not only to AI development and training practices but also in terms of transparency.

This layering of regulation is consistent with the policy judgment of the EU to keep copyright law in itself stable and anthropocentric while AI falls subject to a dedicated regulation of AI systems.

CONCLUSION

The question of ownership and legal recognition for AI-generated works reflects a deep tension in Indian copyright law between its human-centered statutory architecture and the technological fact of creative systems capable of autonomous output production. To illustrate this: at present it is not possible, under the Copyright Act, 1957, to treat AI systems as authors legally, as they fail to be treated as a legal persons and also lack the conscious agency or personal moral investment in creativity that underpins both the economic and moral rights dimensions of copyright. It would be fair to say that with the combined effect of sections 2(d),

17 and 57's statutory text as well as judicial emphasis on human authorship and originality, a

clear schema of excluding non-human entities from copyright authorship, in the absence of any express legislative innovation is established. Yet, statutorily as it stands, little guidance is given to the now frequent situation where human and machine contributions are inextricably intertwined-where a developer designs model architectures, a trainer prepares datasets, an operator sets parameters, a user provides a prompt and the system produces an output through automated processing which each of those actors influence but none fully control Such ambiguity cannot be overcome by Indian courts by interpretation alone, and they need statutory clarification as to how ownership is to be allocated, what level of originality is required from AI-assisted works, and what moral rights apply in respect of such works.

International comparative experience has been instructive on a number of models but offers no ready-made solution. The UK's s 9(3) approach - bestowing copyright on non-human outputs based on a statutory fiction - raises issues of conceptual incoherence and over-protection of routine

machine outputs. The US approach is strictly human authorship trilateral, denies protection to works created solely by AI, maintains conceptual coherence at some under-protection cost to economically valuable outputs, but creates disincentives for AI development. The EU approach applies the traditional standards of originality across the board, supplementing copyright law itself with dedicated AI regulation. This is perhaps a pragmatic but incomplete response to fundamental questions that AI raises about authorship doctrine. The most logical and balanced way ahead for India is legislative amendment that outlines a clear scheme for AI-assisted and AI-generated works.

The reform should (i) retain human authorship as the threshold requirement of copyright, yet define human authorship to include human contributions in the development, training, deployment, and use phases; (ii) make distinctions among types of works based on the extent and nature of human creative input, vesting ownership in humans whose expressive intent and creative judgment are reflected in the output; (iii) retain moral rights under Section 57 as the exclusive rights of humans, making sure those rights remain tied to personality, dignity, and reputation concerns; (iv) set forth clear principles with regard to the liability of developers regarding the compliance of training data and user conduct with respect to prompting AI systems; and (v) require transparency and disclosure norms on the use of AI in creative works to allow downstream users and audiences to make informed decisions with regard to the origin and authenticity of the content.

Such reform would have to be supplemented by judicial development applying these principles to novel factual scenarios, as well as attention to international harmonization, so that Indian law facilitates and does not hinder participation in global digital creative markets. Established by the Indian government in 2025, the committee of experts has a chance to build upon this agenda, and its recommendations should emphasize the necessity of legislative clarity over administrative ad hocism. Only in this way- via constant reform, legislative, judicial, and administrative can India achieve a copyright regime that accommodates AI-generated content while protecting human creators, encourages legitimate innovation, and reflects the more human-centred values copyright law has embodied for a long time.