Authored By: Ayesha Salma Yusoof
Middlesex University Dubai
Abstract
This article examines the scope of the UK’s copyright in the luxury and fashion industry, which has further deepened with the emergence of generative Artificial Intelligence. The article also addresses how Intellectual Property has evolved and how rising concerns about the training and data replication of AI tools pose a risk to the integrity of design and style. Alongside academic and practice-oriented commentary, the dynamics of design protection are primarily analysed through the Copyright, Designs and Patents Act (1988) framework along with landmark cases including Designers Guild Ltd v Russell Williams (Textiles) Ltd (2001) and Response Clothing Ltd v Edinburgh Woollen Mill Ltd (2020). Additionally, the structural gaps relating to authorship of AI-generative designs, unresolved legality and copyright enforceability of AI training systems to protect fashion archives will be interrelated with the current framework and recommendations.
Introduction
Fashion designs sit at a uniquely uncertain intersection of creative expression, commercial capitalism and legal ambiguity. With the global fashion industry generating more than $2 trillion annually, the creative patterns and designs that distinguish a Missoni zigzag from a Burberry tartan remain a commercially valuable form of intellectual property, especially for high-end fashion brands.[1] Despite theoretical protection available to companies in the UK under the Copyright, Designs and Patents Act (1988),[2] the reality for textile designers at the present moment likely reflects an inconsistent judicial application, given the evolving external influences on the fashion industry.
The emergence of generative artificial intelligence design tools like Adobe Firefly and Stable Diffusion enables the production of novel-looking textile designs within seconds of entering a specific prompt. Being trained to generate outcomes based on millions of images from fashion archives and rights-holder directories without license or permission raises questions not only about who owns these generated designs but also about whether these AI training systems are committing copyright infringement. Hence, this article seeks to address whether the UK’s copyright statutory framework is sufficient to protect fashion patterns in an AI-driven age.
The argument advancement not only addresses the statutory framework’s structural failures but also provides a critical comparative analysis and recommendations based on landmark cases observed across jurisdictions.
Background and Conceptual Framework
Under the Copyright, Designs and Patents Act (1988), copyright in terms of fabric patterns would classify as an original artistic work; hence, under Section 4, the definition of artistic work likely encompasses any form of artistic craftsmanship ranging from paintings, drawings and plans.[3] Given that textile prints are fabric decoration, they would therefore fall within the works of graphics. However, structurally crafted fabric designs, which are carefully engineered designs, have a higher and more uncertain threshold in defining them as artistic craftsmanship under section 4 (1) (c).
The UK’s test of originality, as articulated by Justice Peterson in the University of London Press Ltd v University Tutorial Press Ltd (1916), historically referred to originality not being limited solely to novelty but to an independent expression of an individual’s idea, even if they share the same standard of skills, labour or judgment.[4] In contrast, the Court of Justice of the European Union (CJEU) enhanced this standard by referring to originality as the “author’s own intellectual creation” as applied in Cofemel – Sociedade de Vestuário SA v G-Star Raw CV (C 683/17) (2019).[5] However, in the post-Brexit context, the UK would have to reconsider its standards of artistic craftsmanship alongside retained EU law, which could create an unresolved divergence in standards.
With respect to design rights, the UK Unregistered Design Rights, governed under the Copyright, Designs and Patents Act (1988), protects the shape and configuration of original designs within the UK. However, the exclusion of surface decorations under section 213 (2) (c) makes fabric prints and patterns more suited for protection under the Supplementary Unregistered Design Right (SUDR).[6] This mechanism, particularly post-Brexit, provides a three-year protection for surface decorations.[7]
As for AI-generated designs, section 9(3) of the Copyright, Designs and Patents Act (1988) states that the authorship of computer-generated outputs is attributed to the person who undertook the initiative to facilitate the creation.[8] With this clause drafted before the existence of generative AI, such a definition is likely inadequate. Consequently, the UK dual system and limitations on UDR’s reflect the robust protection mechanism of the fashion industry as compared to the EU, which potentially raises concerns about London’s status as a global fashion capital.[9]
Legal Analysis
The central issues relating to the adequacy of copyrights towards protecting textile prints and fabric patterns can be addressed as three distinct challenges: (i) the ownership of generative AI textile designs;(ii) the legality of training AI models to protect fashion archives; (iii) the enforceability of copyrights against AI-generated outcomes which replicate elements of existing protected works.
Authorship of AI-Generated Textile Designs
With section 9(3) of the Copyright, Designs and Patents Act (1988) defining computer-generated authorship, the application of such a statutory definition would mean that the copyright of computer-generated designs lies with the person making the necessary arrangement.[10] With the designer inputting detailed and refining prompts to obtain a desired outcome from the AI, there is a strong argument that the designer is the author despite the AI operating as a creative tool. However, this section fails to address AI autonomous operations. A fast-fashion chain producing multiple textile prints using automated AI processing, without meaningful human creative input, reflects a purely operational arrangement because no human creative decision is involved. Hence, complete ownership is likely void, as AI-assisted designs reward less creative human involvement, rendering it commercially arbitrary.
Additionally, Thaler v Comptroller-General of Patents, Designs and Trade Marks (2023) stated that the possession of intellectual property rights cannot be solely held by the AI system and that the investor or a party holding the rights must be a human entity.[11] Lord Kitchin’s observation is that if AI is used as a sophisticated tool to assist the inventor, it is unlikely to be considered the AI’s own.[12] Therefore, Gucci’s integration of generative AI into its design workflow allows for robust AI usage and substantial human oversight, which greatly contrasts with the idea of a fully automated AI design system in the fast fashion market.[13]
Training AI on Protected Textile Archives
Under section 16 of the Copyright, Designs and Patents Act (1988), copying or reproducing a part of copyright-protected textile images without a license, for purposes like AI training, constitutes infringement.[14] However, implementing such a rule becomes complicated when geographical territory is involved. In Getty Images v Stability AI (2025), the primary infringement claim was effectively dismissed because Stability AI’s training and model developments occurred outside of the UK and would be considered an extraterritorial act not applicable to UK copyright laws.[15] Consequently, this creates an incentive for AI developers to locate training infrastructures offshore to avoid UK copyright liability. Ultimately, the status quo preserves the theoretical copyright liability on UK-based AI training, but training facilities outside the UK are likely to leave independent textile designers and major luxury conglomerates without a workable remedy. Despite the WIPO Hague System allowing worldwide registration of textile designs as part of a licensing framework, the process is likely to be time-consuming and uncertain when protecting fashion archives from industrialisation by the extraterritorial operations of major AI developers.[16]
AI Outputs and the Substantial Part Test
Under the Copyright, Designs and Patents Act (1988), which defines the scope for classifying a work as an original artistic work, the originality standard test and retained EU law relating to the “author’s own intellectual creation” are deemed relevant if the work reflects the author’s independent creative choices rather than technical outcomes. Applying Designers Guild Ltd v Russell Williams (Textiles) Ltd (2001), the introduction of a substantial part test reflects a qualitative assessment of the overall works involving the labour, skills and originality of the author’s work rather than a visual comparison of similarities and differences.[17] Generally, AI’s creative outputs are generated probabilistically and not by direct replication. However, if a model has been specifically trained on a particular protected print, such as Emilio Pucci’s signature paisley, it may generate outputs that are substantially similar in expression but not in style. Therefore, a design that follows the style of Missoni’s zigzags is unlikely to be infringing copyright as compared to a direct replication of Missoni’s specific proportions and colour arrangement in the design. Hence, copyright does not protect style but only the expression of the textile designs.[18]
Case Law Discussion
Designers Guild Ltd v Russell Williams (Textiles) Ltd (2001)
Designers Guild Ltd (DGL) held the copyright of a fabric design “Ixia” which featured flowers and stripes against a brushwork background. Russell Williams Textiles Ltd. (RWT) produced the “Margurite” design, which shared similar structural and compositional characteristics with Ixia. The legal question was whether there was copyright infringement and whether RWT had copied a substantial part of DGL’s copyrighted work. [19]
Lord Hoffmann and Lord Millett held that the two distinct questions:(i) has the defendant copied the work? (ii) was a substantial part copied? found that the Court of Appeal wrongly conflated these by performing a fresh visual comparison at the substantiality stage.[20] This decision established the principle of cumulative originality, where individually commonplace elements like stripes or flowers can be protected if the original outcome reflects the designer’s creative work.[21] The substantial part test is qualitative and susceptible to the interpretation of a reasonable person, making it heavily dependent on the trial judges’ findings, which could create uncertainty. This authority remains a fundamental case on copyright infringement of fashion designs within the UK. It signals that courts will protect the holistic originality of textile designs even if the defendant’s work differs in details and if some elements of the output are not unique.
Response Clothing Ltd v Edinburgh Woollen Mill Ltd (2020)
Response Clothing Ltd had supplied to Edingburgh Woollen Mill Ltd (EWM) with ladies’ tops made with a wave jacquard fabric from 2009 to 2012. The distinctive feature of this fabric was that it was woven and not printed. After the agreement ended, EWM sourced copies of these wave designs from Chinese suppliers, for which Response sued them for copyright infringement, as the wave fabric was a work of artistic craftsmanship.[22] The legal question was whether these woven fabrics could be considered a work of artistic craftsmanship under s.4 (1) (c) of the Copyright, Designs and Patents Act (1988).[23]
Upon HHJ Hacon applying the Bonz Group v Cooke (1994) (NZ) test, which requires the author to be a craftsman and an artist, the court found that the wave design satisfied the criteria as the designer had created and communicated the pattern with an aesthetic intention.[24] Therefore, the loom used to weave the design is considered a means of production and not the creative agent, making the artistic element of creating the design the driving factor for Response to gain copyright ownership. This decision is significant as it extends copyright protection for machine-made (woven) textiles and highlights the requirements for an artistic quality of a textile to be considered for protection. This case also reflects the UK courts’ cautious navigation through the Post-Brexit transition, with the CJEU’s Cofemel decision expanding the test by requiring originality as a requirement.[25] With the Cofemel case’s judgement likely to be retained post the transition, the proof of artistic craftsmanship is likely to be greater for AI-assisted designs given the instructive nature of the design process.
Critical Analysis
The doctrine of artistic craftsmanship imposes a higher and more judicially discretionary threshold on fabric design as compared to other categories like literary and graphic work. Given the age of generative AI, the vague statutory definition of computer-generated output and its ownership will remain more complicated and likely inadequate in the future if not addressed clearly in legislation. Additionally, the effect of Brexit raises concerns about London’s competitiveness as a global fashion capital compared to French and Italian designers, who are likely to have more straightforward trade and fewer copyright complexities.[26]
As for the territorial gap in AI training liability, locating training infrastructure in a jurisdiction with permissive or absent copyright regulation creates a structural incentive for AI developers. The absence of a transparency obligation for AI developers means that rights holders are unlikely to be able to enforce their rights since they are unable to identify whether their work has been used for AI training purposes. Consequently, rights-holders of original designs, particularly independent designers and small and medium- sized fashion houses, bear the disproportionate cost of their creative designs being exploited. Even if action is taken, pursuing multi-jurisdictional litigation against AI training, or even the process of internationally registering the design for copyright under the Hague System, is both expensive and time-consuming.
Conclusion
In summary, the Copyright, Designs and Patents Act (1988), despite providing a theoretical protection for textile design, is seen to have an inconsistent and discretionary standard for resolving the ownership of AI-generated designs. The unresolved legality of AI training on fashion archives, combined with the absence of a statutory transparency obligation on AI developers, highlights the lack of an enforceable mechanism against AI-generated outputs. Therefore, statutory transparent obligations requiring AI developers to disclose training datasets would likely allow designers to enforce their rights and protect the originality of their creative output. Additionally, implementing a mandatory licensing framework, where creative rights holders are remunerated for using their designs in commercial AI training, would help reduce AI exploitation of creative designs. Despite rapid technological developments, it is crucial to recognise and preserve the value of the fashion industry’s creative heritage whilst ensuring that AI-generated designs do not undermine its profitable status within the global fashion market.
Reference(S):
[1] ’10 Key Fashion Industry Statistics for 2025′ (Aeon, 6 November 2025) < https://project-aeon.com/blogs/10-key-fashion-industry-statistics-for-2025 > accessed 6 June 2026.
[2] Copyright, Designs and Patents Act (1988).
[3] Copyright (n2), s.4 (2).
[4] University of London Press Ltd v University Tutorial Press Ltd (1916) 2 Ch. 601, 609.
[5] Cofemel – Sociedade de Vestuário SA v G-Star Raw CV (C 683/17) (2019), para 29.
[6] Copyright (n2), s.213 (2)(c).
[7] ‘Unregistered Design Rights in the UK Explained’ (Sprintlaw, 1 December 2025) <https://sprintlaw.co.uk/articles/unregistered-design-rights-in-the-uk-explained/ > accessed 6 June 2026.
[8] Copyright (n2), s.9(3).
[9] Kaila A Hill, ‘Intellectual Property Frameworks in the United Kingdom: Growth and Challenges Facing London’s Fashion Industry’ (2025) 16(2) Journal for Global Business and Community <https://jgbc.scholasticahq.com/article/144599-intellectual-property-frameworks-in-the-united-kingdom-growth-and-challenges-facing-london-s-fashion-industry > accessed 6 June 2026.
[10] Copyright (n8).
[11] Thaler v Comptroller-General of Patents, Designs and Trade Marks (2023) UKSC 49, para.26.
[12] Thaler (n11), para.52.
[13] ‘8 Ways Gucci is Using AI [Case Study][2026]’ (DigitalDefynd Education, 22 December 2025) < https://digitaldefynd.com/IQ/ways-gucci-using-ai/ > accessed 6 June 2026.
[14] Copyright (n2), s.16 (1).
[15] Getty Images v Stability AI (2025) EWHC 2863 (Ch), para.9.
[16] ‘Benefits’ (World Intellectual Property Organisation) < https://www.wipo.int/en/web/hague-system/benefits > accessed 6 June 2026.
[17] Designers Guild Ltd v Russell Williams (Textiles) Ltd (2001) 1 All ER 700, 704.
[18] Ibid, 705.
[19] Ibid, 703.
[20] Ibid, 704.
[21] Ibid, 706.
[22] Response Clothing Ltd v Edinburgh Woollen Mill Ltd (2020) EWHC 148, paras 2-6.
[23] Copyright (n2), s.4(1) (c).
[24] Response (n22), para 64.
[25] ‘EU copyright in designs – CJEU rule in Cofemel that “originality” is the only requirement for protection’ (CMS, 7 October 2019) < https://cms.law/en/gbr/legal-updates/EU-copyright-in-designs-CJEU-rule-in-Cofemel-that-originality-is-the-only-requirement-for-protection > accessed 6 June 2026.
[26] Kaila A Hill (n9).
