This article is authored by Priyansh Kanawat, a fourth year B.A. LL.B. (Hon’s) student at the National Law University and Judicial Academy, Assam.

India’s geographical indication framework for handloom textiles is architecturally precise in what it protects and structurally silent on what it does not. The Geographical Indications of Goods (Registration and Protection) Act 1999 (“GI Act 1999”) confers rights anchored to geographic provenance and prohibits false origin claims with considerable doctrinal rigour.[1]

The GI Act 1999 draws its multilateral foundation from Article 22 of the Agreement on Trade-Related Aspects of Intellectual Property Rights (“TRIPS Agreement”), supplemented since India’s 2024 accession by the Geneva Act of the Lisbon Agreement on Appellations of Origin and Geographical Indications. [2] Section 22(1) of the GI Act 1999 prohibits the deployment of a registered indication on non-originating goods;[3] Section 22(2) extends that prohibition to any expression or sign likely to suggest false geographic origin.[4] The Madras High Court, in Geofilms & Ors v Geographical Indications Registry,[5] confirmed that Section 22(2) reaches indirect evocations of origin and is not confined to explicit mislabelling. Both provisions are, however, activated exclusively by claims about geographic origin, whether express or implied.[6] A manufacturer producing fabric that precisely replicates the kinkhwab brocade patterns, zari work, and meenakari ornamentation of Banarasi textiles, while labelling it accurately as ‘Banarasi-inspired, made in Surat,’ commits no actionable wrong under the GI Act 1999. This research undertakes a doctrinal analysis of that enforcement vacuum and the instruments surrounding it.

II. The Protection Architecture and Its Structural Limit

India’s textile GI registry encompasses over eighty registrations.[7] Banarasi brocades and sarees, [8] Chanderi fabric,[9] Kanjivaram silk,[10] Pochampally ikat, [11] Phulkari embroidery,[12] Kullu shawls,[13] and Pashmina[14] are among the most commercially significant. The GI Act 1999 draws its multilateral foundation from Article 22 of the Agreement on Trade-Related Aspects of Intellectual Property Rights.[15] Section 22(1) of the GI Act 1999 prohibits deployment of a registered indication on non-originating goods.[16] Section 22(2) extends that prohibition to any expression or sign likely to suggest false geographic origin.[17] The Madras High Court in Geofilms & Ors v Geographical Indications Registry[18] confirmed that Section 22(2) reaches indirect evocations of origin and is not confined to explicit mislabelling. Protection therefore operates robustly against a manufacturer labelling power-loom fabric as ‘Banarasi silk.’

The critical limitation is that both provisions are activated exclusively by claims about geographic origin, whether express or implied.[19] A manufacturer producing fabric that precisely replicates the kinkhwab brocade patterns, zari work, and meenakari ornamentation of Banarasi textiles, while labelling it accurately as ‘Banarasi-inspired, made in Surat,’ commits no actionable wrong under the GI Act 1999.[20] The Act provides no cause of action against aesthetic displacement untethered from false origin representation.[21]

III. The AI Imitation Mechanism

Generative AI systems capable of producing textile designs operate through large multimodal models trained on digitised image corpora.[22] A model trained on archival images of GI-protected textiles acquires the capacity to generate novel designs reproducing the geometric grammar, colour relationships, and ornamental logic of those textiles without copying any specific work.[23] The outputs are novel in the copyright sense and accurate in the origin sense. They nevertheless capture the distinctive aesthetic value of GI-protected products and displace them commercially.[24]

Three structural features of GI-protected Indian handloom textiles intensify their vulnerability. First, their commercial value is substantially constituted by their aesthetic vocabulary rather than solely by geographic provenance.[25] A Kanjivaram saree commands its price premium because it carries the temple-border weave patterns and korvai technique associated with Kanchipuram; an AI-generated fabric replicating those elements captures a significant portion of that premium without originating from the registered region.[26] Second, the registered producers of these textiles, predominantly artisan weaver cooperatives operating through bodies such as the Uttar Pradesh Handloom Fabrics Marketing Cooperative Federation,[27] lack the litigation resources to pursue novel enforcement theories against AI developers. Third, the training data problem adds a further dimension of harm: a model trained without authorisation on digitised GI-protected textile archives may have committed an infringing act during training,[28] while producing outputs that no existing Indian IP instrument can reach.[29] The Office of the Development Commissioner for Handlooms has documented a measurable decline in GI-protected handloom sales since 2023, correlating with the proliferation of AI-generated and AI-assisted power-loom imitations.[30]

IV. The Inadequacy of Existing Instruments

The Copyright Act 1957 does not supply a remedy. Section 15(2) extinguishes copyright in any design applied industrially more than fifty times.[31] Handloom textile designs produced at commercially viable volumes almost invariably exceed that threshold.[32] The Supreme Court of India confirmed this principle in Microfibres Inc v Girdhar & Co,[33] holding that artistic works applied industrially at sufficient scale lose copyright protection entirely, a ruling the Delhi High Court applied in Ritika Private Limited v Biba Apparels Private Limited[34] in the specific context of textile design. The Designs Act 2000 does not supply a remedy either. Design protection requires proof of copying and does not extend to independently created similar designs.[35] A generative AI model producing designs resembling GI-protected textile motifs through statistical inference from a training corpus, rather than by directly copying a registered work, does not infringe a design right under any established doctrine.[36]

Suggested amendments seek to introduce an unregistered design right for novel designs for three years from first disclosure, modelled on Council Regulation (EC) No 6/2002.[37] These reforms are material. They do not, however, address AI-generated independent creation, training data liability, or the intersection of design protection with GI-protected aesthetic vocabularies.[38]

V. The Case for Targeted GI Act Amendment

Three amendments to the GI Act 1999 would close the enforcement vacuum. First, Section 22 should prohibit the commercial use of any design system substantially derived from or replicative of the distinctive aesthetic character of a registered GI-protected textile, regardless of whether the goods carry or evoke the registered indication.[39] This formulation transposes to GI law the reputational protection extended to well-known trademarks under Section 2(zg) of the Trade Marks Act 1999,[40] as developed by the Delhi High Court in Tata Sons Ltd v Manoj Dodia & Ors.[41] It enables enforcement against aesthetic displacement without requiring proof of any false origin representation.[42] Second, the GI Act 1999 should provide that the use of GI-protected textile designs as training data for commercial AI systems requires the authorisation of the registered indication’s authorised users.[43] This is consistent with India’s obligations under Article 39 of the TRIPS Agreement[44] and with the Draft National Intellectual Property Rights Policy 2024.[45] Third, the GI Registry should be empowered to maintain a digitised reference archive of registered textile design vocabularies, structured analogously to the Traditional Knowledge Digital Library.[46] Such an archive would provide AI developers with constructive notice of protected design systems and supply an evidentiary foundation for infringement analysis.[47]

VI. Conclusion

The GI Act 1999 does not protect what constitutes the commercial core of GI-protected handloom textiles: their distinctive aesthetic vocabularies. Generative AI exploits that gap precisely, producing commercially viable designs that displace GI-protected products without triggering any cause of action under existing law.[48] Targeted amendment of the GI Act 1999, extending protection to aesthetic displacement, regulating AI training data use, and establishing a digital reference archive, is both doctrinally grounded and urgently necessary.[49] The traditions of Varanasi, Kanchipuram, and Pochampally were not built to serve as training corpora. Indian intellectual property law must reflect that fact.

---

[1] Dwijen Rangnekar, ‘The International Protection of Geographical Indications: The Asian Experience’ (UNCTAD-ICTSD, 2004) 3.

[2] Geographical Indications of Goods (Registration and Protection) Act 1999 (India).

[3] ibid s 22.

[4] Nayomi Mendez Andrade, ‘Threads of Protection: Intellectual Property Protections for Fashion Designs in the Age of Generative AI’ (2026) 21 Washington Journal of Law, Technology and Arts 1, 6.

[5] Abhilasha Praveen, ‘The House of AI: Intellectual Property Law as it Relates to Generative AI and Fashion’ (2026) 13(1) Pursuit: The Journal of Undergraduate Research at the University of Tennessee 1, 12.

[6] Copyright Act 1957 (India).

[7] GI Registry (India), ‘List of Registered Geographical Indications’ (2026) https://ipindiaservices.gov.in/GIRPublic accessed 24 August 2026.

[8] GI Registry (India), ‘Application No 99: Banaras Brocades and Sarees’ (2009).

[9] GI Registry (India), ‘Application No 7: Chanderi Fabric’ (2005).

[10] GI Registry (India), ‘Application No 15: Kancheepuram Silk’ (2005).

[11] GI Registry (India), ‘Application No 4: Pochampally Ikat’ (2004).

[12] GI Registry (India), ‘Application No 27: Phulkari’ (2010).

[13] GI Registry (India), ‘Application No 19: Kullu Shawl’ (2005).

[14] GI Registry (India), ‘Application No 1386: Ladakh Pashmina Textile’ (2024).

[15] Agreement on Trade-Related Aspects of Intellectual Property Rights (adopted 15 April 1994, entered into force 1 January 1995) 1869 UNTS 299, art 22.

[16] GI Act 1999, s 22(1).

[17] ibid s 22(2).

[18] Geofilms & Ors v Geographical Indications Registry (2010) Madras High Court, WP No 14781 of 2009.

[19] Ministry of Commerce and Industry (India), ‘Report of the Expert Committee on Geographical Indications’ (2018) para 4.3.

[20] ibid para 4.5.

[21] Rangnekar (n 1) 12.

[22] Ryan Abbott and Bret Bogenschneider, ‘Should Robots Pay Taxes? Tax Policy in the Age of Automation’ (2018) 12 Harvard Law and Policy Review 145.

[23] Mendez Andrade (n 4) 8.

[24] ibid 10.

[25] Lotika Varadarajan, A Brocade of Threads: The Textile Heritage of India (Marg Publications 2009) 47.

[26] ibid 52.

[27] Uttar Pradesh Handloom Fabrics Marketing Cooperative Federation Ltd, ‘Registered Producers under the Banarasi Brocades GI’ (UPICA, 2022).

[28] Getty Images (US) Inc v Stability AI Ltd, No 1:23-cv-00135-UNA (D Del, filed 3 February 2023), Amended Complaint paras 48-63.

[29] Department for Promotion of Industry and Internal Trade, Working Paper on Generative AI and Copyright (Part I) (Ministry of Commerce and Industry, 2025) para 5.7.

[30] Office of the Development Commissioner for Handlooms, Ministry of Textiles (India), ‘Annual Report 2024-25’ (2025) 38-41.

[31] Copyright Act 1957 (India) s 15(2).

[32] Praveen (n 5) 14.

[33] Microfibres Inc v Girdhar & Co (2009) 40 PTC 1 (Del HC).

[34] Ritika Private Limited v Biba Apparels Private Limited (2016) 232 DLT 378 (Del HC).

[35] Sudhir Ravindran and Vinita Radhakrishnan, ‘Does India Need an Unregistered Design Right?’ (2023) 18 Journal of Intellectual Property Law and Practice 410, 415.

[36] Ravindran and Radhakrishnan (n 35) 417.

[37] Council Regulation (EC) No 6/2002 of 12 December 2001 on Community Designs [2002] OJ L3/1, art 11.

[38] Working Paper (n 29) para 5.9.

[39] Kal Raustiala and Christopher Jon Sprigman, ‘The Piracy Paradox: Innovation and Intellectual Property in Fashion Design’ (2006) 92 Virginia Law Review 1687, 1720.

[40] Trade Marks Act 1999 (India) s 2(zg).

[41] Tata Sons Ltd v Manoj Dodia & Ors (2011) Delhi High Court, CS(OS) No 264 of 2008.

[42] P Narayanan, Intellectual Property Law (3rd edn, Eastern Law House 2001) 312.

[43] Working Paper (n 29) para 6.2.

[44] TRIPS Agreement (n 15) art 39.

[45] Department for Promotion of Industry and Internal Trade, National Intellectual Property Rights Policy (Ministry of Commerce and Industry, 2016) objective 5.3.

[46] Council of Scientific and Industrial Research (India), ‘Traditional Knowledge Digital Library’ https://www.tkdl.res.in accessed 24 August 2026.

[47] ibid; Department for Promotion of Industry and Internal Trade (India), ‘Report of the Expert Committee on the Review of the Designs Act 2000’ (DPIIT, 2022) para 7.4.

[48] Mendez Andrade (n 4) 15.

[49] Praveen (n 5) 18.