Algorithmic Copying and the Adequacy of UK Unregistered Design Rights in the Age of AI-Driven Fashion

Abstract

The rapid integration of artificial evidence into fashion deisgn processes has generated significant legal uncertainty within the United Kingdom’s intellectual property framework. This article examines whether the UK’s unregistered design rights regime, which comprises of the UK Unregistered Design Right under the Copyright, Designs and Patents Act 1988 and the Supplementary Unregistered Design Right under the retained Community Design Regulation, provides adequate protection against the twin challenges of AI-assisted design generation and AI-facilitated algorithmic copying by fast fashion actors. Through an analysis of relevant legislation, key judicial decisions, and the UK Intellectual Property Office’s 2025 consultation on the designs framework, this article argues that the current framework’s reliance on human authorship and sequential copying fails to address the realities of algorithmic design generation. Structural legislative reform is urgently required.

Keywords: Generative AI, Algorithmic Copying, Unregistered Design Rights, Supplementary Unregistered Design Right, Originality, UK Intellectual Property.

Introduction

The relationship between artificial intelligence and the fashion industry has evolved rapidly from theoretical curiosity to commercial reality. Luxury houses like Gucci and Tommy Hilfiger have deployed generative AI systems trained on brand archives to develop new silhouettes and fabric patterns, while ultra-fast fashion platforms, notably Shein, have attracted sustained litigation alleging that proprietary algorithms systematically identify and reproduce designs of independent creators at a scale and speed no human copying operation could replicate. The commercial stakes are considerable, with Shein’s annual revenues having exceeded thirty billion US dollars in 2023.[1] This is a figure built in substantial part upon an algorithmic infrastructure capable of generating approximately ten thousand new product listings per day.

The legal frameworks through which designers have historically sought to protect their work were not designed with algorithmic generation in mind. In the United Kingdom, unregistered design rights offer automatic protection without the cost or delay of formal registration, making them of particular importance to independent and emerging designers who lack the resources to engage the registered design system. Yet these rights carry inherent limitations, as they are narrow in scope, brief in duration and conditional upon proof of copying. This exposes them to fundamental inadequacy in an era of AI-driven design generation and copying.

This article argues that the UK’s unregistered design rights framework fails on two distinct fronts: it creates doctrinal uncertainty around the protection of designs produced with AI assistance, and it presents near-insurmountable evidentiary barriers when enforcing against algorithmic copying by fast fashion operators. The article proceeds by setting out the relevant legal framework before applying it to the AI context through an analysis of key case law, concluding with a critical assessment of the framework’s adequacy and the reform options currently under consideration by the UK Intellectual Property Office.

Background: The UK Unregistered Design Rights Framework

The protection of fashion designs in the United Kingdom rests upon two distinct unregistered regimes that coexist in a relationship of complementary but imperfect coverage. The UK Unregistered Design Right (‘UDR’), established by sections 213 to 264 of the Copyright, Designs and Patents Act 1988 (‘CDPA’), protects the shape or configuration of three-dimensional articles.[2] It arises automatically upon creation, subsists for a maximum of fifteen years from first recording or ten years from first marketing, and critically does not extend to surface decoration. It is therefore of limited relevance to the most commercially significant elements of fashion design, namely the surface patterns, textile prints, and ornamental features that characterise garments as aesthetic objects rather than functional structures.

This gap is addressed, albeit incompletely, by the Supplementary Unregistered Design Right (‘SUDR’), which was introduce into domestic law following the United Kingdom’s departure from the European Union to replicate the protection formerly afforded by the Unregistered Community Design under the Community Design Regulation (EC) No 6/2002.[3] The SUDR protects the overall appearance of a product, including its shape, colours, texture, materials and ornamentation, and subsists for three years from the date on which the design is first disclosed to the public.[4] For fashion designers, the SUDR is the primary unregistered right of practical relevance, as most garment designs concern visual appearance rather than structural engineering.

Both regimes are subject to threshold requirements of originality and, in the case of SUDR, individual character. The design must produce an overall impression on the informed user that differs from that produced by any design previously made available to the public. Fundamentally, both regimes presuppose a human author. The CDPA’s provisions on computer-generated works, found at section 9(3), attribute authorship to the person who makes the necessary arrangements for the creation of a work where no human author exists, but this provision applies to copyright works rather than to design rights, and its extension to the design context remains legally uncertain. As the following analysis demonstrates, this uncertainty is the source of the framework’s most significant inadequacies.

The Authorship Problem in AI-Assisted Design

The most acute doctrinal challenge posed by generative AI to the unregistered design rights framework is the question of authorship. The concept of originality, as developed through UK and EU jurisprudence, in inherently human-centric. The Court of Justice of the European Union established in Infopaq International A/S v Danske Dagblades Forening that a work is protectable only if it reflects the author’s own intellectual creation, requiring the author to have exercised free and creative choices to imprint their personal touch.[5] This standard was confirmed as applicable in the UK post-Brexit context by the Court of Appeal in THJ Systems Limited v Sheridan [2023], which affirmed that the originality threshold requires the author to have exercised sufficient skill and creative judgment.[6]

At first glance, this requirement appears to exclude AI-generated outputs from protection entirely. Where a designer employs a generative adversarial network and issues an open-ended prompt, the fundamental expressive features of the output are determined probabilistically by the AI system rather than by the human operator. However, this framing overstates the problem for many designers that deploy AI as a creative tool. It is submitted that, correctly analysed, the Infopaq standard accommodates AI-assisted design provided the human designer’s creative contribution is identifiable across the generative process as a whole.

There are three stages at which a designer may exercise the requisite free and creative choices. First, in the curation of training data: a designer who selects specific archival images, fabric references, aesthetic parameters to fine-tune an AI model steers the model’s generative outputs toward a particular creative vision, and this editorial act itself reflects creative judgment. Second, in prompt engineering: the specification of aesthetic parameters and stylistic properties thorugh carefully constructed prompts requires the exercise of creative choice arguably no less than the specification of a brief to a human studio. Third, in output selection and modification. The designer who reviews a set of machine-generated outputs, selects the most promising, and applied modifications before the design reaches production is exercising precisely the kind of curatorial and transformative judgment that the personal touch doctrine is designed to protect. Niyompatama and Lapatoura, writing in the Journal of Intellectual Property Law and Practice, have convincingly argued that generative AI functions as a complementary design tool and that designers who exercise creative control across these stages retain sufficient authorship to meet the originality threshold.[7]

The critical qualification is evidentiary rather than doctrinal. The right may exist in principle but establishing it in practice demands that designers document their creative process with a rigour that the current framework does not require and that most design workflows do not naturally produce. The difficult is compounded by the probabilistic nature of AI outputs. Where a prompt is open-ended or a model is overtrained on pre-existing designs, establishing that a particular output reflects the designer’s personal touch rather than the model’s statistical tendencies may be impossible without transparency into the AI system itself. This evidentiary burden falls asymmetrically on smaller designers who lack the resources to maintain detailed records of AI-assisted workflows.

A sharper and more intractable problem arises in relation to fully AI-generated designs, where outputs produced with minimal human input beyond the activation of a system. The claim to authorial creative choice is considerably more strained, and where the existing framework with absent legislative clarification may offer no protection at all. The distinction between AI-assisted and fully AI-generated designs is therefore not merely academic. It determines whether the framework’s inadequacy is a matter of evidentiary difficulty or doctrinal exclusion. Both demand different legislative responses.

The Individual Character Requirement and Algorithmic Outputs

A further challenge concerns the individual character requirement for SUDR protection. AI systems trained on large fashion archives that potentially may hold thousands of prior designs, may produce outputs that closely resemble, or constitute probabilistic recombination of pre-existing designs without precisely reproducing any single one. The UK Intellectual Property Office’s 2025 consultation on the designs framework has explicitly acknowledged this risk, noting that AI could produce a massive volume of designs in a very short time, potentially flooding the market and making it harder for human-created designs to meet originality requirements.[8] This is not merely a theoretical concern, as if the proliferation of AI-generated designs renders the prior art landscape so dense that human designs cannot establish individual character, the practical effect would be to systematically disadvantage human designers in favour of those with access to AI infrastructure.

Enforcement Against Algorithmic Copying

The third and most commercially pressing dimension concerns the enforcement of unregistered design rights against operators that employ AI for systematic copying. Both UDR and SUDR require the claimant to establish that the defendant copied the protected design. This is a causal requirement that has proven difficult to satisfy even in the context of traditional human-led fast fashion. In the context of algorithmic copying, this burden becomes considerably more onerous.

A platform may deploy an algorithm that identifies trending designs across social media platforms and transmits outputs directly to manufacturers without human review. This is the model alleged in Perry, Martinez and Baron v Shein.[9] The causal connection between the original design and the infringing product is mediated by a machine process whose internal workings are proprietary and opaque. Establishing copying under UK unregistered design law requires evidence that goes beyond the mere fact of similarity, creating a structural barrier to enforcement that algorithmic opacity significantly compounds.

Case Law Discussion

The most significant authority on the scope of unregistered design protection in the fashion context is Karen Millen Fashions Ltd v Dunnes Stores, in which the Court of Justice of the European Union was asked to clarify the test for individual character under the Community Design Regulation.[10] The dispute arose from Dunnes Stores’ admitted copying of two Karen Millen garments, following which Dunnes contended that the designs lacked individual character, arguing that their component features could be found, in combination, among earlier designs.

The CJEU rejected this ‘mosaicking’ argument, holding that the individual character of a design must be assessed against one or more specific prior designs taken individually, rather than against a composite of features drawn from several earlier designs.[11] This ruling carries significant implications for the AI context. First, it clarifies that a defendant cannot defeat a design right claim merely by demonstrated that individual design features are each commonplace, with the assessment being holistic. Second, and by implication, an AI system capable of generating outputs that differ in their overall impression from any single prior design, while recombining elements from many, might produce designs that satisfy the individual character requirement, notwithstanding the absence of genuine human creative expression. The case remains persuasive authority in UK courts following Brexit.

Response Clothing Ltd v The Edinburgh Woollen Mill Ltd is the most directly relevant recent UK authority on unregistered design rights in the fashion context.[12] Response Clothing had supplied garments featuring a wave-pattern fabric to Edinburgh Woollen Mill, which subsequently sourced similar garments from alternative suppliers following a commercial dispute. The Intellectual Property Enterprise Court held that while the wave fabric did not qualify as a graphic work under section 4(1)(a) of the CDPA, it constituted a work of artistic craftsmanship under section 4(1)(c).[13] The judge applied the Bonz test, requiring that the creator intended to produce a work of art that also served a functional purpose.

In Cofemel, the CJEU held that garments may attract copyright protection in the EU where they constitute the author’s own intellectual creation, rejecting any additional requirement for aesthetic quality as a condition of protection.[14] The decision directly influenced the reasoning in Response Clothing, in which the court applied the EU-derived originality standard. Although, Cofemel is no longer binding in the UK following Brexit, it remains persuasive authority. Its affirmation that originality, rather than aesthetic merit, is the primary threshold for protection clarifies that the barrier to protection in the AI context lies not in evaluating the quality of a design but in establishing the authorial creative process. If the threshold for protection is purely originality in the Infopaq sense, AI-generated designs are excluded not because they lack beauty or commercial appeal but because they cannot be attributed to human creative expression.[15]

Critical Analysis

An assessment of the UK’s unregistered design rights framework in the context of AI-driven fashion reveals three structural inadequacies that are not addressable through judicial interpretation alone and that demand legislative attention.

First, the framework offers an uncertain and evidentially onerous basis for protecting AI-generated or substantially AI-generated designs. The human authorship requirement, embedded in the originality standard through Infopaq and THJ Systems, means that fully generated AI-generated outputs may generate no protection at all. This creates a perverse commercial consequence whereby brands that invest in AI infrastructure to accelerate design generation may find that the resultant designs are unprotected, while brands that document human creative involvement may retain rights. The law thereby creates incentives for the manufacture of paper trails of human input rather than genuine creative engagement. The IPO’s 2025 consultation has acknowledged this directly, noting that the originality and individuality character requirements are inherently human-centric concepts and that it is not clear how they apply when a design is created entirely by a machine.[16]

Second, the SUDR’s three-year duration is structurally inadequate in the context of algorithmic fast fashion. A design identified as trending by an AI system and algorithmically copied may reach retail within weeks of first public disclosure. Three years of protection may appear sufficient in isolation, but the combination of that limitation with the evidentiary burden of proving copying narrows the window for effective enforcement to the point of practical futility. This is compounded by the opacity of proprietary algorithmic processes. The Law Society, in its response to the IPO’s 2025 consultation, noted that small businesses and independent designers are particularly reliant on unregistered rights precisely because they cannot afford registration, yet these designers are the most vulnerable to algorithmic copying and the least able to sustain complex IP litigation.[17]

Third, the framework contains no mechanism for addressing the systemic dimension of algorithmic copying. Individual design right claims address individual garments. They do not address the algorithmic infrastructure producing thousands of infringing products simultaneously. Academic commentary has drawn attention to the need for a remunerative solution, such as AI developers who use in-copyright fashion designs for commercial training would need to pay a statutory fee to original creators.[18] This would act to reconcile the interests of IP holders and AI developers without prohibiting legitimate use. The prospect of criminal sanctions for unregistered design infringement, under consideration in the IPO’s 2025 consultation, may represent a necessary step toward addressing algorithmic copying at the systemic level.[19]

Conclusion

The UK’s unregistered design rights framework was constructed for a world in which designers were human, copying was sequential, and infringement could be identified through direct comparison of products. The rapid integration of artificial intelligence into both design generation and design copying has rendered each of these assumptions problematic. The human authorship requirement creates doctrinal uncertainty around the protection of AI assisted designs. The evidentiary demands of proving copying present near-insurmountable barriers when the alleged infringement is algorithmic and opaque. The three-year duration of the SUDR is also inadequate when copying can be executed at industrial scale within days of a design’s first appearance.

The IPO’s 2025 designs framework consultation represents a timely recognition of these inadequacies, and its exploration of reforms to the originality requirement, the consolidation of the two unregistered regimes, and the potential introduction of criminal sanctions for infringement are all germane to the challenge posed by AI.[20] However, effective reform will require not merely technical adjustments to existing rights but a fundamental reconsideration of whether a framework built upon human authorship and individual copying is adequate for the commercial realities of AI-driven fashion. Legislative intervention that addresses AI authorship, mandates algorithmic transparency in infringement proceedings, and strengthens enforcement mechanisms is both necessary and, given the pace of technological development, urgent.

Bibliography

Primary Sources – Legislation

Copyright, Designs and Patent Act 1988

Community Design Regulation (EC) No 6/2002

Intellectual Property Act 2014

Registered Designs Act 1949

Primary Sources – Cases

Cofemel – Socidedad de Vestuário SA v G-Star Raw CV (C-683/17) EU:C:2019:721, [2019] Bus LR 2796.

Infopaq International A/S v Danske Dagblades Forening (C-5/08) EU:C:2009:465, [2009] ECR I-6569.

Karen Millen Fashions Ltd v Dunnes Stores, Dunnes Stores (Limerick) Ltd (C-345/13) EU:C:2014:2013, [2014] Bus LR 1163.

Perry, Martinez & Baron v Shein Distribution Corp et al (C.D. Cal., Case No 2:23-cv-05551, filed July 2023).

Response Clothing Ltd v The Edinburgh Woollen Mill Ltd [2020] EWHC 148 (IPEC), [2020] FSR 25.

THJ Systems Limited & Anor v Daniel Sheridan & Anor [2023] EWCA Civ 1354.

Secondary Sources – Journal Articles

Geiger C and Iaia V, ‘Fashion, Intellectual Property and Freedom of Artistic Expression in the Age of Metaverse and AI’ [2024] 46(9) European Intellectual Property Review.

Niyompatama MJ and Lapatoura I, ‘Generative AI in fashion design creation: a copyright analysis of AI-assisted designs’ [2025] 20(10) Journal of Intellectual Property Law & Practice.

Other Sources

Intellectual Property Office, ‘Consultation on changes to the UK designs framework’ (GOV.UK, 2025) https://www.gov.uk/government/consultations/consultation-on-changes-to-the-uk-designs-framework/consultation-on-changes-to-the-uk-designs-framework  accessed 6 June 2026.

The Law Society, ‘Intellectual Property Office Consultation on the UK Designs Framework: Response’ (The Law Society, 2026) https://www.lawsociety.org.uk/campaigns/consultation-responses/intellectual-property-office-consultation-on-uk-designs-framework accessed 6 June 2026.

The Fashion Law, ‘A New Lawsuit Lifts the Lid on Shein & its AI-Powered Model’ (TFL, 2024) https://www.thefashionlaw.com/a-new-lawsuit-lifts-the-lid-on-sheins-ai-powered-ultra-fast-fashion-model/ accessed 6 June 2026.

[1] The Fashion Law, ‘A New Lawsuit Lifts the Lid on Shein and its AI-Powered Model’ (The Fashion Law, 2024) https://www.thefashionlaw.com/a-new-lawsuit-lifts-the-lid-on-sheins-ai-powered-ultra-fast-fashion-model/ accessed 6 June 2026.

[2] Copyright, Designs and Patents Act 1988, ss 213-264.

[3] Community Design Regulation (EC) No 6/2002, art 11; Intellectual Property (Unjustified Threats) Act 2017, s 1.

[4] Community Design Regulation (EC) No 5/2002, art 11(1).

[5] Infopaq International A/S v Danske Dagblades Forening (C-5/08) [2009] ECR I-6569

[6] THJ Systems Ltd v Sheridan [2023] EWCA Civ 1354, [2024] ECDR 4

[7] Mark Jetsaphon Niyompatama and Ioanna Lapatoura, ‘Generative AI in fashion design creation: a copyright analysis of AI-assisted designs’ (2025), 20(10) Journal of Intellectual Property Law & Practice, 654.

[8] Intellectual Property Office, ‘Consultation on changes to the UK designs framework’, (GOV.UK, 2025), https://www.gov.uk/government/consultations/consultation-on-changes-to-the-uk-designs-framework accessed 6 June 2026, para 11.

[9] Perry v Shein Distribution Corp (2023) No 2:23-cv-05551 (CD Cal)

[10] Karen Millen Fashions Ltd v Dunnes Stores Case C-345/13, EU:C:2014:2013

[11] ibid

[12] Response Clothing Ltd v The Edinburgh Woollen Mill Ltd [2020] EWHC 148 (IPEC)

[13] ibid

[14] Cofemel – Sociedade de Vestuário SA v G-Star Raw CV Case C-683/17, EU:C:2019:721

[15] Infopaq International A/S v Danske Dagblades Forening (n 5)

[16] IPO Consultation (n 8), paras 11 and 16.

[17] The Law Society, ‘Intellectual Property Office Consultation on the UK Designs Framework: Law Society Response’ (The Law Society, 2026) https://www.lawsociety.org.uk/campaigns/consultation-responses/intellectual-property-office-consultation-on-uk-designs-framework  accessed 6 June 2026.

[18] Christopher Geiger and Vincenzo Iaia, ‘Fashion, Intellectual Property and Freedom of Artistic Expression in the Age of Metaverse’ (2024), 46(9) European Intellectual Property Review 542.

[19] IPO Consultation (n 8).

[20] Law Society Response (n 17)