Authored By: Sanyameka Bhardwaj
Amity University, Noida
Abstract
The integration of generative artificial intelligence into fashion design has outpaced the law that is supposed to govern it. This article examines the specific and underexplored crisis created for Indian intellectual property law when a fashion design is produced, not by a designer working with digital tools, but by an AI system operating with substantial autonomy. Drawing on the Copyright Act, 1957, and the Designs Act, 2000, the article maps the fault lines in existing doctrine: the ambiguous phrase ’causes the work to be created’ in Section 2(d)(vi), the originality standard established in Eastern Book Company v D B Modak and its fraught application to non-human outputs, the structural tension between copyright and design registration under Section 15, and the absence of a dedicated ownership framework for any of the several possible claimants — the designer who prompts, the developer who trained the model, or the enterprise that deployed it. The article further considers the December 2025 DPIIT Working Paper and its conspicuous silence on the authorship question. It proposes a ‘creative control’ test as an interpretive solution within the existing statute, and argues that, in the longer term, India should develop a sui generis framework tailored to the realities of the fashion industry rather than simply borrowing the United Kingdom’s computer-generated works model. The dispute is in relation to India’s particular industrial conditions a textile sector of global impact, a large informal design market, and a tradition that has historically ranked human creativity over others.
Keywords: Generative AI; fashion design; Copyright Act 1957; Section 2(d)(vi); Designs Act 2000; authorship; originality; sui generis; DPIIT Working Paper 2025; computer-generated works.
I. Introduction
In January 2024, a menswear label debuted a collection at a London trade fair in which every initial sketch draft had been proposed by Midjourney, the generative AI tool, from a series of text prompts written by the brand’s sole designer. The designer had then selected, adapted, and executed the chosen outputs into physical garments. The label owned physical clothes. Nobody was entirely sure who, if anyone, owned the underlying designs.
This is not an abstract question. The Indian fashion and textiles industry, one of the country’s most economically significant sectors, is undergoing a rapid and largely unregulated encounter with generative AI. Tools such as Midjourney, Adobe Firefly, Stable Diffusion, and purpose-built platforms like AiDLab and Yoona.ai are now embedded in design workflows across industry. [3]According to Adobe’s own Digital Trends data, more than sixty-eight per cent of creative teams globally were using AI features to generate or modify visual content by 2025[4]. At the luxury end, LVMH’s Director of Brand Protection reported submitting 2.5 million counterfeit-content reports to platforms in 2024 alone, attributing a material share of that volume to AI-facilitated infringement[5]. The creative and commercial stakes are, therefore, substantial in today’s time.
Indian IP law is not equipped to resolve these stakes. The Copyright Act, 1957 [6]and the Designs Act, 2000 [7]were designed, in their fundamental architecture, around the assumption that creative works are produced by human beings. The 1994 amendment to the Copyright Act introduced Section 2(d)(vi), which provides that the author of a computer-generated work is ‘the person who causes the work to be created.[8]‘ This was a forward-looking provision in its day, and it provides the most obvious statutory hook for addressing AI authorship in India. But it was drafted for a world in which a human programmer caused a computer to execute defined operations — not for a world in which a diffusion model, trained on hundreds of millions of images, produces a novel textile print in response to a three-word prompt.
The December 2025 Working Paper published by the Department for Promotion of Industry and Internal Trade (DPIIT) [9]acknowledges the authorship problem on its face, identifying ‘the copyrightability of AI-generated works’ and ‘identifying authorship in AI-generated content’ as matters requiring resolution. [10]But Part I of the Working Paper, which is the portion so far published, does not resolve them. The Bar and Bench has noted that under the current statute, ‘AI systems do not presently fall within the definition of an author under Indian law.[11]
This article seeks to fill that gap, with a specific focus on fashion design — a creative domain that has received virtually no India-specific academic attention in this context. One recent paper published in early 2026 identified the gap at the level of general IP law; [12]none, to the present author’s knowledge, has addressed the specific intersection of AI authorship, the copyright/design boundary under Section 15, and the practical commercial questions that arise when an autonomous system produces garment designs at scale.
The article proceeds as follows. Part II situates the problem by describing the technological context and its implications for the fashion industry. Part III examines the authorship question under the Copyright Act, analyzing Section 2(d)(vi) and the originality threshold from Eastern Book Company. Part IV turns to the Designs Act and the critical overlap problem under Section 15. Part V assesses the DPIIT’s December 2025 position and the comparative model offered by the UK’s Copyright, Designs and Patents Act 1988. Part VI proposes the ‘creative control’ test as an interpretive solution within current law. Part VII argues for a long-term sui generis framework suited to India’s conditions, and Part VIII concludes.
II. The Technological Context: Generative AI in the Fashion Design Workflow
To reason clearly about authorship, it is necessary to understand what generative AI actually does in a design context, and how substantially it can differ from older forms of computer-assisted creation.
The term ‘generative AI’ consists of a range of various large language models, generative adversarial networks (GANs), variational autoencoders, and, most prominently in the visual domain, diffusion models. These systems are trained on datasets of enormous scale: Midjourney and Stable Diffusion, for example, were trained on hundreds of millions of images drawn from the internet, including, the works of professional fashion designers, illustrators, and textile artists. Training teaches the model the statistical relationships between visual features and the human language used to describe them. Once trained, the model can generate novel images from text prompts with no further human creative input at the point of output.
Within the fashion industry, three distinct modes of AI use can be identified, and they raise meaningfully different legal questions.
The first, and most legally tractable, is AI as a sophisticated creative tool. Here, a human designer writes detailed prompts, iterates across many outputs, selects particular results, and then makes substantial modifications — adjusting colorways, redrawing silhouettes, adding surface detail — before arriving at the final design. The AI functions as a kind of advanced generative brush. The designer’s creative choices are pervasive. This mode is broadly analogous to a photographer choosing angle, light, and moment, and it presents the strongest case for attributing authorship to the designer.[13]
The second mode is AI as autonomous creative agent. A fashion enterprise feeds a brief ‘geometric prints suitable for a summer kurta collection inspired by Madhubani painting’ — to an AI system, which produces thirty candidate designs. The enterprise selects five and sends them to a manufacturer. The human contribution is confined to (a) commissioning the brief, (b) selecting among outputs, and (c) deciding to commercialize. Whether that degree of human involvement constitutes ‘causing the work to be created’ within the meaning of Section 2(d)(vi) is, as this article argues, genuinely uncertain.
The third mode, which is becoming increasingly prevalent, is fully automated AI design generation for fast-fashion production. Here, an enterprise deploys a trained model to continuously produce new print designs, which are automatically routed to a cut-and-make supply chain with no individual human selecting any particular design. Human involvement is limited to the initial configuration of the system. This mode sits furthest from any conventional notion of authorship and closest to industrial manufacture.[14]
The legal significance of these distinctions is that Indian IP law has no established mechanism for treating them differently. Section 2(d)(vi) applies the same test ‘causes the work to be created’ to all three. The article returns to this problem in Part VI.
III. Authorship Under the Copyright Act, 1957
A. The Definition of ‘Author’ and the Causation Paradox
The Copyright Act defines ‘author’ in Section 2(d), which sets out a category-specific list. [15]Section 2(d)(iii) provides that, in the case of an artistic work, the ‘author’ is the artist. Section 2(d)(vi), introduced by the Copyright (Amendment) Act, 1994, provides that in relation to a computer-generated work, the author is ‘the person who causes the work to be created.’[16]
The drafting history of Section 2(d)(vi) is instructive. The 1994 amendment was introduced primarily to address the position of computer software and, more broadly, to adapt the Act to digital creation. [17]The provision was influenced by, though not identical to, the formula adopted in the United Kingdom’s Copyright, Designs and Patents Act 1988 (‘CDPA’), which locates authorship in ‘the person by whom the arrangements necessary for the creation of the work are undertaken.’[18] It is broader because it does not limit authorship to whoever makes ‘arrangements’; it extends to any person in the causal chain whose acts were a cause of the work’s creation. It is more uncertain because causation, in a legal and philosophical sense, is famously difficult to delimit.
The causation paradox in the context of generative AI may be stated simply. When Midjourney produces a textile print derived from a prompt, multiple parties arguably ’caused’ the work to be created as their work was already taken into consideration as that information is fed and by default referred to produce output. The designer who wrote the prompt caused the output design, in the sense that without the prompt, this particular output would not have existed. The developer who built and trained the model caused it, in the more fundamental sense that without the model, no output of this type would have been possible at all. The enterprise that licensed and deployed the model caused it, in an organizational sense. And the hundreds of thousands of visual artists and textile designers whose works formed the training data caused it, in the aggregated sense that the model’s visual vocabulary is derived entirely from their creative output.
Indian courts have not yet been required to resolve this paradox in any decided case[19]. In the absence of judicial guidance, the academic literature has identified two principal interpretive approaches. The first, favored by commentators who prioritize continuity with the existing scheme, is to treat ’causes the work to be created’ as requiring proximate human causation — meaning that authorship should vest in the human whose creative decision-making was the most direct cause of the specific output. [20]The second approach, less consistent with the Act’s text but more responsive to the commercial reality of AI deployment, is to treat the provision as admitting any person or entity that bears legal responsibility for the act of creation — which, in a corporate deployment scenario, would typically be the employing entity.[21]
Neither approach is satisfactory as a complete solution. The first collapses when applied to the third mode of AI use described in Part II — fully automated generation — because there is no human whose creative decision-making was the proximate cause of any individual design. The second, as currently argued, requires a degree of purposive construction that the text of Section 2(d)(vi) probably cannot bear without legislative intervention.
B. The Originality Threshold After Eastern Book Company
Even if the authorship question could be resolved in favour of a human claimant, a further obstacle arises from the originality requirement. Copyright in an artistic work subsists under Section 13 of the Act only if the work is ‘original.’ [22]The leading Indian authority on the content of originality is the Supreme Court’s decision in Eastern Book Company & Ors v D B Modak & Anr, [23]which rejected both the bare ‘sweat of the brow’ doctrine and the more demanding American standard, settling instead on a ‘middle path’ requiring that the author demonstrate skill and judgment that go beyond the mechanical.
The application of this standard to AI-generated fashion design is problematic in either direction. Where a designer has written a detailed, iterative series of prompts and made substantive selections among AI outputs, the argument that the designer’s skill and judgment are reflected in the final design is plausible. The Supreme Court in Eastern Book Company indicated that even modest exercises of editorial judgment could suffice, [24]and there is a reasonable case that a skilled prompt engineer, making deliberate aesthetic choices, brings equivalent judgment to the process.
The difficulty intensifies, however, as human contribution becomes more attenuated. In the second and third modes of AI use, human’s contribution is closer to commissioning a brief and selecting a product than to exercising creative skill. The argument that such involvement satisfies the Eastern Book Company test is weak. And if the originality threshold is not met, no copyright subsists at all — the design enters the public domain on creation, providing no protection against copying. The commercial consequences for fashion enterprises that have invested significantly in AI-driven design workflows would be severe.
There is a further wrinkle. The Eastern Book Company test assumes that the skill and judgment in question are those of a human author. It is, in principle, no different from the question asked in Feist Publications [25]— does the putative author bring sufficient independent creative thought to the work? Where creative thoughts is entirely the AI’s, there is no author in whose skill and judgment originality can be located. The work simply does not fit the statutory scheme as currently constructed.
C. The Ownership Question: Who Gets the Right?
Assuming, for the sake of argument, that a human claimant can be identified as ‘author’, the first ownership of copyright vests in that author under Section 17 of the Act, subject to the provisos. [26]In the employment context, copyright vests in the employer where the work is made in the course of employment under a contract of service. In a commissioned design context, the Act provides no equivalent first-ownership rule for commissioners (unlike, for example, UK law in certain categories), so that copyright would vest in the author unless contractually assigned.
In practice, the service agreements of AI tool providers (Midjourney’s, Adobe Firefly’s, and Stable Diffusion’s terms of service all address outputs to varying degrees) purport either to vest output ownership in the user or, in some cases, to retain a license. Indian contract law would govern the enforceability of these terms, but since no Indian court has addressed the intersection of these contractual arrangements with the copyright statute, the legal position is uncertain at this point, and no definite statutes are set for now.
The problem of the training data creators — the original artists and designers whose works trained the model — adds a further dimension. In Ani Media (P) Ltd v Open AI Inc [27]the Delhi High Court has been confronted with a claim that AI training on copyrighted journalistic content constitutes infringement. The outcome of that litigation may have significant implications for the unanswered question “whether AI-generated outputs are themselves derivative works that infringe the copyright in training data”, but no final judgment had been delivered at the time of writing.
IV. The Designs Act, 2000 and the Section 15 Problem
A. Design Protection Without a Human Designer
The Designs Act, 2000 provides a potentially complementary route to protection for fashion designs. A ‘design’ under Section 2(d) of that Act means features of shape, configuration, pattern, ornament, or composition of lines or colors applied to an article — precisely the kind of output that a generative AI model produces when tasked with creating a textile print or a garment silhouette. [28]The Act requires the design to be ‘new or original’, not previously disclosed to the public, and significantly distinguishable from known designs.[29]
There is no express provision in the Designs Act, 2000 that restricts registration to designs created by human beings. The Act defines neither ‘author’ nor ‘creator’, and its provisions dealing with the proprietor of a design focus on the applicant rather than on the originator of the design. On a literal reading, therefore, it might be argued that a human enterprise could register an AI-generated design under the Designs Act and thereby obtain a monopoly, provided that the design is new, original, and not excluded on other grounds.
However, this interpretation has not been tested before the Indian registrar or any court, and there are conceptual objections to it. The Act’s framework of rights and remedies presupposes a human or legal-person proprietor capable of exercising and enforcing those rights. More practically, the ‘novelty’ and ‘originality’ requirements would need to be assessed against a global prior-art base that now includes the outputs of AI systems themselves — an assessment that current examination procedures are not designed to perform at scale.
B. The Section 15 Structural Tension
The most practically significant complexity in the copyright/design interface for AI-generated fashion designs arises from Section 15 of the Copyright Act. [30]Section 15(1) prevents copyright from subsisting in any design registered under the Designs Act. Section 15(2) provides that copyright in any design ‘capable of being registered’ under the Designs Act ceases once the design has been reproduced more than fifty times by an industrial process.
The consequences for AI-generated fashion design are potentially severe. An AI-generated textile print that is applied to a garment and reproduced at scale — which is precisely how fast fashion operates — will almost certainly satisfy the ‘capable of being registered’ test, triggering the Section 15(2) deadline. Once fifty reproductions have been made, copyright is extinguished. If the enterprise has not registered the design under the Designs Act before that point, it will have no protection at all.
Under conventional human-authored design practice, the Section 15(2) mechanism, while commercially important, is navigable through planning. Designers who expect to reproduce their work industrially can register under the Designs Act before reaching the fifty-reproduction threshold. But in a fully automated AI design pipeline — the third mode described in Part II — designs may be produced and reproduced at a rate that makes this registration procedure practically infeasible. A model generating hundreds of print designs per day, with manufacture commencing almost immediately, cannot realistically be accompanied by a design registration for each output.
The result is a structural mismatch between the pace of AI-driven fashion production and the procedural architecture of Indian design law. This is not merely a practical inconvenience; it creates a legal environment in which investment in AI-driven design generation carries significant unprotected risk — a consequence that neither the Copyright Act nor the Designs Act was designed to produce and that the law should, if it is to be coherent, address.[31]
C. The Artistic Work/Industrial Design Boundary
A further dimension of the Section 15 problem is the distinction, developed in Indian jurisprudence, between fashion designs as works of genuine artistic expression (attracting full copyright) and fashion designs as commercially motivated industrial designs (attracting only the more limited design-law protection).
The Fashion Law Journal has observed that ‘runway designs that showcase originality and artistic flair can be safeguarded as works of art through copyright protection’ while ‘fast fashion, driven entirely by commercial motives, won’t find refuge under the Copyright Act in India.’ [32]This binary is, of course, imprecise — the line between artistic expression and industrial design is notoriously difficult to draw even for human-authored works — but it acquires additional complexity when the author is an AI. Can an AI-generated fashion design ever be said to express genuine artistic intent? The question has both philosophical and legal dimensions that are not resolved by existing doctrine.
V. The DPIIT Working Paper and the Comparative Question
A. The December 2025 Working Paper: An Assessment
The DPIIT’s Working Paper on Generative AI and Copyright, published on 8 December 2025, is [33]a significant policy document and a welcome acknowledgement that India’s copyright framework requires reconsideration in light of generative AI. It constitutes, as one commentator has described it, ‘India’s first serious attempt to construct a regulatory framework for AI training on copyrighted works.’[34]
The Working Paper’s principal proposals concern the input side — that is, the use of copyrighted works as training data for AI models. It proposes a hybrid licensing model, providing that all lawfully accessed copyrighted content should be available for AI training as a matter of right, subject to a statutory remuneration scheme for copyright owners. [35]This is a reasonable response to the training-data problem and broadly consistent with approaches being developed in the European Union and the United Kingdom.
However, for the purposes of this article, the more important observation is that the Working Paper’s treatment of the output side — the authorship and copyrightability of AI-generated works — is conspicuously incomplete. Part I of the Working Paper flags the questions of ‘copyrightability of AI-generated works’ and ‘identifying authorship in AI-generated content’ as matters to be addressed. [36]But it explicitly defers those questions to the forthcoming Part II. Part II had not published as of the date of writing this article.
The legislative and policy vacuum is therefore ongoing. As the Bar and Bench has noted, ‘AI systems do not presently fall within the definition of an author under Indian law. [37]Any enterprise in India that designs with AI operates in a condition of genuine legal uncertainty about who, if anyone, owns the resulting designs.
B. The UK Model: Lessons and Limitations
The United Kingdom’s Copyright, Designs and Patents Act 1988 is the most developed statutory framework for computer-generated works anywhere in the world, [38]and Indian IP scholarship has naturally looked to it for guidance. Section 9(3) of the CDPA provides that, where a literary, dramatic, musical or artistic work is computer-generated ‘in circumstances such that there is no human author of the work’, the author is ‘the person by whom the arrangements necessary for the creation of the work are undertaken.’ [39]Section 178 defines a ‘computer-generated’ work as one ‘generated by computer in circumstances such that there is no human author’. [40]The protection term is reduced to fifty years from making[41], and the author has no moral rights.[42]
At first glance, this framework appears to offer a ready-made solution for the Indian context. Transpose Section 9(3) into Indian law, identify the ‘person who undertakes the arrangements’ as the author, and the problem is — at least in structural terms — resolved.
The analysis is, however, more complicated. Several objections apply.
First, Section 9(3) has never been tested in the English courts on facts resembling modern generative AI. The provision was drafted in 1988 for a conception of ‘computer-generated’ work that involved human-authored programs executing defined operations — not large language models producing outputs that their own creators cannot predict. Academic commentators have noted that the provision ‘has never been tested in court’ and that there is an ‘inherent problem’ in its mechanism because it requires an ‘LDMA work’, which presupposes sufficient originality even while locating authorship in the person who makes arrangements rather than a human creator.[43]
Second, the ‘arrangements’ test is not obviously superior to India’s ’causes the work to be created’ formulation. Both locate authorship in a person or entity external to the AI, and both leave open the question of how to identify that person when multiple parties (prompt engineer, enterprise licensee, model developer) have each contributed to the arrangements.
Third, and most importantly from an Indian perspective, a transplant approach risks importing a framework whose implicit assumptions are calibrated to the UK’s creative economy and legal culture, without adequate attention to the specific features of the Indian fashion and textiles industry. The Berne Convention, to which India is a party, permits member states to determine the conditions of their own copyright law, including the definition of authorship, provided that minimum standards of protection are maintained. [44]India is therefore free to develop its own approach.
VI. The Creative Control Test: An Interpretive Proposal
Within the existing text of Section 2(d)(vi), there is interpretive room for a more principled approach than either of the two positions identified in Part III. This article proposes what it calls the ‘creative control test’ as a framework for determining, under current law, who ’causes the work to be created’.[45]
The test asks: at the point of the work’s creation, which identifiable human being exercised substantive and directive creative control over the specific aesthetic choices embodied in the work? ‘Substantive’ control means that the person made genuine creative decisions — about the visual qualities, the aesthetic direction, the stylistic choices — that are reflected in the output. ‘Directive’ means that the person’s choices were causally determinative of the specific output produced. ‘Specific aesthetic choices’ means the choices that distinguish this design from any other design that could have been produced.
Applied to the three modes of AI use identified in Part II, the test produces the following results. In the first mode (AI as tool), the designer who wrote the detailed prompts, selected among outputs, and made iterative aesthetic decisions satisfies the test and is the author under Section 2(d)(vi). This is consistent with the Eastern Book Company standard of skill and judgment.
In the second mode (AI as autonomous agent with human selection), the outcome is less certain. Where the human’s contribution was confined to writing a general brief and selecting among AI-generated outputs, the question is whether selection alone — choosing one design from thirty — constitutes substantive and directive creative control over the specific aesthetic choices in the selected design. This article argues that it does not, at least where the selector has not materially modified the output. The act of selection reflects taste and commercial judgment, but it does not reflect the skill and judgment contemplated by Eastern Book Company. On this analysis, no copyright would subsist in the selected design under the second mode, which has important commercial implications.
In the third mode (automated generation), no human satisfies the creative control test, and Section 2(d)(vi) cannot be made to produce an author. The designs fall outside the copyright scheme entirely.
The creative control test would need to be further specified through legislative intervention or judicial elaboration, and it has limitations. It does not resolve the position of the AI developer or the training data creators. And it does not address the structural problem under Section 15 of the Copyright Act, which operates independently of the authorship question. But as an interpretive framework for Section 2(d)(vi) within the existing statute, it is more principled and more commercially coherent than either of the available alternatives.
The test also has implications for the Section 17 first-ownership rules. [46]Where the designer who exercises creative control is an employee acting in the course of employment, copyright vests in the employer in the usual way. Where the designer is an independent contractor commissioned by a fashion enterprise, copyright would vest in the contractor unless contractually assigned. These are workable results, consistent with the general scheme of the Act.
VII. The Case for a Sui Generis Indian Framework
A. Why the Current Statute Cannot Do Enough
The creative control test proposed in Part VI is an interpretive solution to the question of who authors a computer-generated work under existing law. It is not a comprehensive solution to the policy problem. As the foregoing analysis makes clear, there are at least three distinct gaps that interpretive creativity cannot fill.
First, the automated generation scenario (the third mode) produces no author and no copyright. A legislature that wishes to provide some protection for the outputs of automated AI design systems — or that wishes to ensure those outputs enter the public domain with clarity — must say so expressly. Leaving the question unanswered creates uncertainty that serves no one’s interests.
Second, the Section 15 problem — the 50-reproduction threshold and the interaction between copyright and design registration — is a structural tension that no interpretation of the authorship provisions can resolve. It requires either amendment of Section 15 or the creation of a new registration procedure suited to the pace of AI-driven production.
Third, the interests of the training data creators — the original artists and textile designers whose works trained the AI models — are not addressed by any of the currently available doctrinal frameworks. This is the problem that DPIIT’s hybrid licensing model attempts to address on the input side, but the output side remains unexplored.
B. What a Sui Generis Framework Should Contain
A sui generis framework for AI-generated fashion designs in India should, in the present author’s view, address five matters.
The first is a clear definition of ‘AI-generated design’, distinguishing it both from ‘AI-assisted design’ (where a human exercises creative control throughout) and from conventional computer-generated work. The distinction should turn on the degree of human creative control at the moment of generation, using a version of the creative control test as the operative criterion.
The second is a limited, time-restricted neighboring right in AI-generated designs — not copyright, but an exclusive right of exploitation for a shorter period, perhaps fifteen years. This would provide a commercial incentive for investment in AI design tools without creating the kind of long-term monopoly that copyright would confer and without implying that AI outputs embody the kind of human creativity that copyright was historically designed to reward.
The third is a simplified registration procedure for AI-generated designs that can operate at the scale at which such designs are produced. The current Designs Act registration procedure is manual, fee-bearing, and examination-dependent. An automated bulk-registration mechanism — perhaps supplemented by a publicly accessible registry of AI-generated design outputs — would provide a workable solution to the Section 15 problem.
The fourth is a transparency and disclosure obligation, requiring that AI-generated designs be identified as such when used commercially. This is consistent with the approach of the EU AI Act, [47]and would serve the interest of human designers in not having their work falsely attributed to human creative effort.
The fifth is a statutory mechanism for compensating training data creators — the artists and designers whose works trained the AI models — out of a proportion of the revenues generated by AI-designed fashion products. DPIIT’s hybrid licensing model gestures in this direction but has not yet specified how it would operate for the fashion design sector specifically.
C. The Indian Context as a Distinctive Consideration
India’s textile and fashion industry is one of the largest in the world and employs tens of millions of people, many of them weavers, embroiderers, and artisans working in informal and semi-formal settings. [48]The traditional craft designs of these workers — block prints, ikat patterns, Madhubani motifs — are precisely the kind of visual material on which generative AI models are trained. A framework that permits AI enterprises to commercialize designs derived, even if statistically rather than directly, from these traditions without any mechanism of compensation or attribution would be both legally anomalous and culturally unjust.
India’s constitutional framework, which recognizes the protection of traditional knowledge as a matter of public interest, provides a basis for legislating in this space that goes beyond the narrowly proprietary concerns of copyright law. The Geographical Indications of Goods (Registration and Protection) Act, 1999 is one existing instrument that gestures in this direction, but it does not extend to the AI context. A comprehensive AI and fashion design framework should consider whether the interest of traditional craft communities in their design vocabularies should be protected by a right that is collective, perpetual, and specifically calibrated to the AI training context.
This argument is not merely sentimental. It is a recognition that the specific conditions of the Indian textile economy — the geographic concentration of craft traditions, the economic vulnerability of informal artisans, the global commercial significance of Indian design vocabularies — create a set of interests that are neither addressed by the UK’s CDPA model nor contemplated by the DPIIT’s current Working Paper.
VIII. Conclusion
When the loom runs itself — when a generative AI model autonomously produces a textile print that enters commercial production without meaningful human creative involvement — India’s intellectual property law does not, at present, have a coherent answer to the question of who owns the output. Section 2(d)(vi) of the Copyright Act, 1957 was designed for a world in which a human programmer directed a computer to execute defined operations. It is not equipped, without judicial elaboration or legislative intervention, to resolve the paradoxes of autonomous AI creation.
The originality threshold from Eastern Book Company — which asks for skill and judgment beyond the mechanical — proves difficult to satisfy where the human contribution is confined to writing a brief or selecting among AI outputs. The Section 15 mechanism, which ties copyright protection to a fifty-reproduction threshold, creates a structural mismatch with the pace of AI-driven fashion production that cannot be resolved by interpretation alone. DPIIT’s December 2025 Working Paper has acknowledged the problem on its face but deferred its resolution.
This article has proposed, as an immediate interpretive solution within the existing statute, a ‘creative control test’ for the application of Section 2(d)(vi): authorship should vest in the human who exercised substantive and directive creative control over the specific aesthetic choices embodied in the work at the moment of its creation. This test is consistent with the Eastern Book Company standard, workable within the existing ownership framework under Section 17, and calibrated to the range of modes in which AI is actually used in the fashion design workflow.
In the longer term, however, the creative control test is not enough. India should develop a sui generis framework for AI-generated fashion designs that provides a time-limited neighbouring right for automated AI outputs, a simplified registration procedure suited to scale, a transparency obligation requiring disclosure of AI generation, and — most distinctively Indian in character — a compensatory mechanism for the traditional artisans and craft communities whose design vocabularies have been absorbed into the training data of global AI systems. The UK model offers a starting point but not an endpoint. The specific conditions of the Indian fashion and textiles industry — its economic significance, its craft traditions, its global reach, and its informal economy — justify and require a legislative response that is distinctively India’s own.
The loom is running. The law needs to decide what to do about it.
Reference(S):
TABLE OF STATUTES, CASES AND OTHER MATERIALS
Primary Indian Legislation
The Copyright Act, 1957 (No. 14 of 1957, as amended)
The Copyright (Amendment) Act, 1994 (No. 38 of 1994)
The Copyright (Amendment) Act, 2012 (No. 27 of 2012)
The Designs Act, 2000 (No. 16 of 2000)
The Geographical Indications of Goods (Registration and Protection) Act, 1999 (No. 48 of 1999)
Indian Cases
Eastern Book Company & Ors v D B Modak & Anr, (2008) 1 SCC 1
Ani Media (P) Ltd v Open AI Inc, 2024 SCC OnLine Del 8120 (Delhi High Court, pending)
Foreign Legislation
Copyright, Designs and Patents Act 1988 (UK), ss 9(3), 12(7), 79(2)(c), 178, 214, 267
Regulation (EU) 2024/1689 (EU Artificial Intelligence Act)
International Instruments
Berne Convention for the Protection of Literary and Artistic Works 1886 (as revised at Paris, 1971)
Policy Documents
Department for Promotion of Industry and Internal Trade (DPIIT), Working Paper on Generative Artificial Intelligence and Copyright (Part I), Working Paper No. P-24029/34/2025-IPR-VII (8 December 2025)
Secondary Sources — Journal Articles and Book Chapters
E Geiger, ‘Generative AI in Fashion Design Creation: A Copyright Analysis of AI-Assisted Designs’ (2025) 20(10) Journal of Intellectual Property Law & Practice 654 (Oxford Academic)
Mohan Kumar SK, ‘The Legal Status of AI-Generated Fashion in Authorship, Ownership and Design Rights’ (2026) Indian Journal of Law and Legal Research
P Naithani, ‘Issues of Authorship and Ownership in Work Created by Artificial Intelligence: Indian Copyright Law Perspective’ (2022) 11(1) NTUT Journal of Intellectual Property Law & Management 7
A Pokhriyal and V Gupta, ‘Artificial Intelligence Generated Works Under Copyright Law’ (2020) 6(2) NLUJ Law Review 1
A Rahmatian, ‘Originality in UK Copyright Law: The Old Skill and Labour Doctrine Under Pressure’ (2013) 44(1) IIC — International Review of Intellectual Property and Competition Law 4
SF Hedrick, ‘I Think, Therefore I Create: Claiming Copyright in the Outputs of Algorithms’ (2019) 8 New York University Journal of Intellectual Property & Entertainment Law 324
Online and Other Sources
Bar and Bench, ‘India’s Generative AI Moment: Copyright Law at Regulatory Crossroads’ (23 January 2026) https://www.barandbench.com/columns/indias-generative-ai-moment-copyright-law-at-regulatory-crossroads
Mondaq, ‘Indian Fashion Industry: Intersection of Copyright and Design Law’ (16 July 2024) https://www.mondaq.com/india/copyright/1493464
Mondaq, ‘Generative AI & Copyright Law in India: Who Owns Machine-Made Works?’ (21 July 2025) https://www.mondaq.com/india/patent/1653344
A Shearman, ‘Ownership of AI-Generated Content in the UK’ (July 2025) https://www.aoshearman.com/en/insights/ownership-of-ai-generated-content-in-the-uk
Fashion Law Journal, ‘Protecting Fashion in India: Supreme Court Revisits the Copyright–Design Overlap’ (20 May 2025) https://fashionlawjournal.com
FashionNetwork India, ‘What Legal Challenges Does the Fashion Industry Face in the Age of Generative AI?’ (10 December 2025) https://in.fashionnetwork.com/news/1790466
Record of Law, ‘The Legal Status of AI-Generated Fashion Designs: Ownership Authorship and IP Protection’ (11 April 2026) https://recordoflaw.in
Record of Law, ‘AI-Generated Fashion Designs: Who Owns the Runway’ (13 April 2026) https://recordoflaw.in
[2] BA.LL.B 4th year, Amity University Noida
[3] The Copyright Act, 1957, No. 14 of 1957, as amended by the Copyright (Amendment) Act,1994 and the Copyright (Amendment) Act, 2012 (India).
[4]The Designs Act, 2000, No. 16 of 2000 (India).
[5]The Copyright (Amendment) Act, 1994, No. 38 of 1994 (India), inserting Section 2(d)(vi)into the Copyright Act, 1957.
[6]Copyright Act, 1957, Section 2(d)(vi): ‘in relation to any literary, dramatic, musical or artistic work which is computer-generated, the person who causes the work to be created’.
[7]Copyright Act, 1957, Section 2(p): ‘artistic work’ means- (i) a painting, a sculpture, a drawing (including a diagram, map, chart or plan), an engraving or a photograph, whether or not any such work possesses artistic quality; (ii) a work of architecture; and (iii) any other work of artistic craftsmanship.
[8]Copyright Act, 1957, Section 13(1)(a) read with Section 13(3)(a): original artistic works are protected subject to the condition that they are original.
[9]Eastern Book Company & Ors. v. D.B. Modak & Anr., (2008) 1 SCC 1 (India). The Supreme Court, applying a standard intermediate between ‘sweat of the brow’ and the American ‘modicum of creativity’, held that for a work to qualify as ‘original’, the author must demonstrate skill and judgment beyond the merely mechanical.
[10]Designs Act, 2000, Section 2(d): ‘design’ means only the features of shape, configuration, pattern, ornament or composition of lines or colors applied to any article whether in two dimensional or three dimensional or in both forms … but does not include any artistic work as defined in clause (c) of section 2 of the Copyright Act, 1957.
[11]Copyright Act, 1957, Section 15: (1) Copyright shall not subsist under this Act in any design which is registered under the Designs Act, 2000. (2) Copyright in any design, which is capable of being registered under the Designs Act, 2000 but which has not been so registered, shall cease as soon as any article to which the design has been applied has been reproduced more than fifty times by an industrial process by the owner of the copyright or, with his license, by any other person.
[12]Copyright, Designs and Patents Act 1988 (UK), Section 9(3): ‘In the case of a 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.’
[13]Copyright, Designs and Patents Act 1988 (UK), Section 178: ‘computer-generated’ means that the work is generated by computer in circumstances such that there is no human author of the work.
[14]Copyright, Designs and Patents Act 1988 (UK), Section 12(7): ‘where the work is computer-generated the above provision does not apply and copyright expires at the end of the period of 50 years from the end of the calendar year in which the work was made’.
[15]Ani Media (P) Ltd. v. Open AI Inc., 2024 SCC Online Del 8120 (Delhi High Court). The first major Indian case to judicially consider the liability of an AI company for allegedly using copyrighted journalistic content as training data. The case was pending final adjudication as of the date of writing.
[16]Department for Promotion of Industry and Internal Trade (DPIIT), Working Paper on Generative Artificial Intelligence and Copyright (Part I), Working Paper No. P-24029/34/2025-IPR-VII, dated 8 December 2025 (India).
[17]Ibid. The Working Paper identified, on the ‘output side’, the need to determine the copyrightability of AI-generated works, identify authorship in AI-generated content, and examine the applicability of moral rights.
[18]Bar and Bench, ‘India’s Generative AI Moment: Copyright Law at Regulatory Crossroads’ (23 January 2026) https://www.barandbench.com/columns/indias-generative-ai-moment- copyright-law-at-regulatory-crossroads
[19] 17Mohan Kumar SK, ‘The Legal Status of AI-Generated Fashion in Authorship, Ownership and Design Rights’ (2026) Indian Journal of Law and Legal Research https://www.ijllr.com accessed 3 June 2026. This paper, published in March 2026, is one of the first academic works to identify this precise gap in the context of Indian IP law.
[20]Collina Strada and other labels presented pieces developed with AI-assistance at New York Fashion Week 2024. See Record of Law, ‘The Legal Status of AI-Generated Fashion Designs: Ownership Authorship and IP Protection’ (11 April 2026) https://recordoflaw.in accessed 3 June 2026.
[21]Mohan Kumar SK (n 17). The author notes that AiDLab, Adobe Firefly, Midjourney, Stable Diffusion, Runway, and Yoona.ai are among the most widely deployed generative AI tools within fashion product development workflows, with adoption accelerating markedly from 2024 onwards.
[22]Adobe Digital Trends Report (2025) cited in aivancity.ai, ‘Design: Our Selection of the Best Generative AI Tools of 2026’ (15 February 2026): ‘more than 68% of creative teams are already using AI features to speed up graphic design or explore new visual directions’.
[23]LVMH Director of Online Brand Protection Nicolas Lambert, speaking at the Assises Juridiques de la Mode, du Luxe et du Design (Paris, 9 December 2025), reported submission of 2.5 million reports of counterfeit content to platforms in 2024 alone, noting that generative AI had substantially lowered barriers to producing infringing material. Reported in Fashion Network India (10 December 2025).
[24]Copyright Act, 1957, Section 2(d)(i): in relation to a literary, dramatic, musical or artistic work, means the author of the work; Section 2(d)(iii): in the case of an artistic work, the artist.
[25]Statement of Objects and Reasons, Copyright (Amendment) Act, 1994 (India). The legislative history indicates that the 1994 amendment was designed to adapt the 1957 Act to developments in computer technology and software, and that Section 2(d)(vi) was modelled in part on the approach taken in the United Kingdom’s Copyright, Designs and Patents Act 1988.
[26]P Naithani, ‘Issues of Authorship and Ownership in Work Created by Artificial Intelligence: Indian Copyright Law Perspective’ (2022) 11(1) NTUT Journal of Intellectual Property Law & Management 7.
[27]Indian Journal of Integrated Research in Law, ‘Adapting Indian Copyright Law to the Age of Artificial Intelligence: Recognizing AI as Authors under the Copyright Act of 1957’ (2024) https://ijirl.com accessed 3 June 2026. The paper argues that non-human entities are not expressly barred from authorship under the Act, though this interpretation has not been judicially endorsed.
[28]Eastern Book Company (n 7), para 46. The Court stated: ‘To be original, a work must not only be the independent creation of the author, but must also possess at least some minimal degree of creativity … the requisite level of creativity is extremely low’.
[29]Copyright Act, 1957, Section 17: the author of a work shall be the first owner of the copyright therein, subject to provisos (a)-(e) dealing with, inter alia, works made during the course of employment, works made under contract of service, photographs, portraits, and government works.
[30]Feist Publications Inc v Rural Telephone Service Co, 499 US 340 (1991) (US Supreme Court). Although not binding in India, the Feist test was discussed approvingly in Eastern Book Company (n 7) and has influenced Indian jurisprudence on what constitutes originality.
[31]Copyright Act, 1957, Section 2(d)(vi): the statutory phrase ’causes the work to be created’ does not specify proximity of causation. It is unclear whether the provision requires the causation to be the proximate cause, or whether it extends to any person in the causal chain, including a person who wrote or licensed the underlying training data.
[32]Thaler v Vidal, 43 F.4th 1207 (Fed. Cir. 2022) (US Court of Appeals, Federal Circuit): ‘only a natural person can be an inventor’. See also US Copyright Office Review Board, Re: Second Request for Reconsideration for Refusal to Register a Recent Entrance to Paradise (14 February 2022), refusing to register a work created autonomously by an AI without any human authorship.
[33]Vegap v Mango (2024) (Spain). The case concerned the digitization of in-copyright paintings by Miró, Tapie’s and Barcelo, and their modification into digital NFT fashion wearables. The court held that the derivative fashion creations infringed the reproduction and adaptation rights of the original artists’ estates. Discussed in E Geiger, ‘Generative AI in Fashion Design Creation: A Copyright Analysis of AI-Assisted Designs’ (2025) 20(10) Journal of Intellectual Property Law & Practice 654.
[34]E Geiger, ‘Generative AI in Fashion Design Creation: A Copyright Analysis of AI-Assisted Designs’ (2025) 20(10) Journal of Intellectual Property Law & Practice 654 (Oxford Academic). The article draws a distinction between AI functioning as a tool (where the human designer makes iterative creative choices) and AI operating with substantial independence.
[35]Designs Act, 2000, Section 2(d). A ‘design’ must (i) be new or original, (ii) not be disclosed to the public anywhere in India or in any other country prior to filing, (iii) be significantly distinguishable from known designs, and (iv) not comprise or contain scandalous or obscene matter. There is no express requirement that the creator be human.
[36]Mondaq, ‘Indian Fashion Industry: Intersection of Copyright and Design Law’ (16 July 2024) https://www.mondaq.com/india/copyright/1493464 accessed 3 June 2026: ‘Dual statutory protection under the design and copyright statute is not possible in India’.
[37]Copyright Act, 1957, Section 15(2). The provision has been interpreted by Indian courts as creating a regime in which industrial-scale reproduction of a design effectively converts copyright protection into design-law protection. See the discussion in Mondaq, ‘Overlap Between Copyrights and Designs in India’ (19 August 2022).
[38]Fashion Law Journal, ‘Protecting Fashion in India: Supreme Court Revisits the Copyright- Design Overlap’ (20 May 2025): ‘runway designs that showcase originality and artistic flair can be safeguarded as works of art through copyright protection. On the flip side, fast fashion, driven entirely by commercial motives, won’t find refuge under the Copyright Act in India. Instead, it must look to the Design Act of 2000 for the protection it seeks’.
[39]Copyright, Designs and Patents Act 1988 (UK), Section 79(2)(c): moral rights, including the right of attribution under Section 77, do not apply to computer-generated works.
[40]A Rahmatian, Originality in UK Copyright Law: The Old Skill and Labour Doctrine Under Pressure (2013) 44(1) IIC International Review of Intellectual Property and Competition Law 4. For a critique of the CDPA computer-generated works provisions in modern generative AI context, see A Shearman, Ownership of AI-Generated Content in the UK (July 2025) https://www.aoshearman.com/en/insights accessed 3 June 2026.
[41]DPIIT Working Paper (n 14). The paper flagged ‘the absence of legal exceptions for AI training’ and noted that ‘Indian copyright law currently provides no specific exception for text and data mining or for AI training activities’. It proposed a hybrid regulatory model providing statutory remuneration rights for copyright owners whose works are used in AI training.
[42]Bar and Bench (n 16) stated that AI systems do not presently fall within the definition of an author under Indian law, nor does the working paper propose any provision that attributes authorship of AI-generated works to AI systems.
[43]The proposed ‘creative control test’ is the present author’s own formulation, synthesising the analytical frameworks of Eastern Book Company (n 7), the DPIIT Working Paper (n 14), and comparative analysis of Section 9(3) of the CDPA 1988 (UK). It is not drawn from any existing statutory provision.
[44]Copyright Act, 1957, Section 17, proviso (c): in the case of a work made by an author in the course of his employment under a contract of service or apprenticeship, to the extent that the work was made in the course of such employment, the employer shall, in the absence of any agreement to the contrary, be the first owner of the copyright therein.
[45]Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Article50 imposes transparency obligations requiring that AI-generated content be labelled as such. The EU AI Act came into force on 1 August 2024 and applies in phases through to 2026.
[46]Berne Convention for the Protection of Literary and Artistic Works 1886 (as revised), Article 2(1): protection extends to ‘every production in the literary, scientific and artistic domain’. The Berne Convention does not itself define ‘author’, leaving that question to national law, but its underlying philosophy is premised on human creative agency.
[47]The Indian fashion industry’s economic significance is reflected in the Ministry of Textiles’ National Textiles Policy 2020 and associated export promotion frameworks. The apparel and textiles sector is one of India’s largest employers. The intersection of AI-driven design with this sector thus has implications not only for IP law but for labour, industrial policy, and export competitiveness.
[48]Microfibres Inc v Girdhar & Co, 2006 SCC OnLine Del 540 the leading Indian case on exactly where artistic copyright ends and industrial design begins
