Authored By: Gunjan Ukey
Kalinga University
ABSTRACT
As AI practice in fashion design becomes widespread, difficult questions surrounding the authorship and ownership of AI-created works are arising. Current copyright rules and principles are developed based on human authorship, however generative AI models possess the capacity to generate sketches, pattern designs and even full design pitches with little to no direct human input. This raises challenges to copyright law and principles in respect of authorship and ownership. This paper analyzes the current copyright regime and explores if copyright law offers a suitable framework to assess authorship and ownership of AI-created fashion designs. This paper takes a comparative approach and analyzes legislations, case law and scholarly opinion by inspecting copyright law in India, the UK and the US. This paper asserts that copyright law frameworks are heavily human-centric and poorly suited to classify an AI as an author or co-author.
KEYWORDS: Intellectual property; Artificial intelligence; Design of clothing; Copyright law; Authors’ rights; Title of ownership; Works of authorship; Works made by or for AI
INTRODUCTION
AI‘s vast affect could also be illustrated in fashion designing. Today, AI systems are able to process historical clothing trends, client profiles, textile libraries and techniques and suggest new styles. In addition to designers, brands and individual professionals are already exploring generative AI to stimulate their inspiration in fashion sketching and modeling, process and technique generation, and color scheme definition, up to a full collection. However, this development accelerates, and amplifies designs but compromises existing intellectual property rights. Works of human intellectual creation are copyrightable. There is a presumption which is likely implicit that traditional ideas about authorship, originalities and ownership of creation relies on an agentive human author who creatively made aesthetic decisions and had total control over the completed work. When an AI works un-aided by human hand, producing a garment that was an autonomous creation of the machine, doubts arise as to the author of the work as defined for the purposes of copyright. Is copyright held by the programmer of the AI or the user of the AI, or the fashion house to which the copyright will ultimately be licensed, or indeed by no one at all? The uncertainty in law also means there is another important point, as innovation, brand value, commercial value within fashion, etc all rely upon the existing IP system. Established Indian, UK and US systems do offer differing solutions to originality and authorship issues but they do not solve the problems posed by advanced generative AIs. This paper intends to answer that question: How much, if any, copyright and ownership does a human fashion designer now have over a generative AI fashion design, and what legal reform is needed for these issues. In answer to this question, this article will start with an analysis of the conceptual and legal basis of the term “author” under copyright.
BACKGROUND AND CONCEPTUAL FRAMEWORK
Authorship is the foundation of copyright. Historically copyright has been attributed to original works authored by a person (human being). Copyright standards at international level such as Berne Convention for the Protection of Literary and Artistic Works implies human author without specifically defining what is human authorship. Most national copyright legislations were developed based on natural person‘s authorship. In the fashion industry, the artistic aspect of copyright mainly involves textile pattern & design, illustration and some artwork. Designs, of course, are perfectly suited to be protected by design rights, trade marks and trade dress, but the extensive use of AI to generate fashion content poses difficulties to such protection. Generative AI‘s operate by ‘feeding’ enormous amounts of data into them and then seeking structure from existing works. These systems, through the implementation of machine learning, have the ability to create art-like and apparently real content. Traditional generating devices were an apparatus for human artist and designers, but leading generative models have the capacity of selecting themselves according to their design quality and features. The debate about the copyrightability of AI created works will tend to touch on three of the main theories of authorship. First is the human authorship model which gives a copyright only when there is human authorship where human contribution is more than minimal. Secondly is the programmer authorship model where the programmer owns the program. The third model is the user ownership model where the user is considered the owner if he/she is the one who started it by giving instructions to the program. Means, for instance, as available under s9(3) of the Copyright Designs and Patents Act 1988 (specific to the UK system) can be used as a legal device in this respect by creating the “author” the person who commissioned the assembly of a computer generated work. On the other hand, the US Copyright Office also invariably states that the copyright can only apply to works generated by a human author. Analogous to Indian copyright law, human intelligence is also assumed under Indian copyright law. The variety of such options already highlights key problems that arise for the copyright regime with advancing AI. As no international copyright regime of AI has been developed, some designers and their owners, designers and fashion firms are not allowed to license their creations to the market.
LEGAL ANALYSIS
The advent of generative AI has had a disruptive effect on copyright. All systems of copyright have been established on the basis that creations emanate from the human thought process. However, the creation of fashion design by AI poses a threat to this because the process is not one involving human thought.
Human Authorship as a Copyright Requirement
Most copyright systems still adhere to the principle of human authorship: Most of the time, copyrighted material are original works that need some level of “intellectual effort”. In fashion, the originality lies in the clothes, the choice of style (shape, design, way of using the material, colours used and appearance). Things become even more tricky with the case of generative AI systems: they are capable of creating their own designs based on a minimum of instructions. If a customer has the idea “a sustainable luxury evening dress influenced by the Mughal Empire architecture,” designers may release designs that reflect concepts that are completely unknown and out of the customers understanding and control. As this makes one question who is the true author.
USA remains the strongest.
Copyright Office of the United States states that “A work must have been subjectively created by a human author in order to qualify for copyright protection”. Thus, if a machine “automatically generated” the expressive elements of a design for the user, it is not copyrightable. It “preserves” the “original intention of the copyright law” but poses difficulty for the companies engaged in the design of fashion items. Section 9(3) CDPA 1988 is approached with a slightly more liberal view in the UK. States that as far as computer program generated work, as well as computer generated work, author is the person responsible for the initiating activities that led to creation of work. This rule existed pre modern AI, and then was later added in as a co-Collaborative rule in Comp 201. However, such rule may be applicable in case of AI-created fashion too. The final product could be referred as designer, clothing company or even company which developed the process. But then still so many unknown. There is also inadequate definition of what is “necessary arrangement”. In the case of processes carried out by several individuals: a software engineer, a data provider, a machine trainer and an end-user, to whom would originality be attributed?
Position under Indian Copyright Law
Indian Copyright Act 1957; There is no direct mention of AIG. With respect to 2(d), it focuses on authors (it is a human author implied). Indian decisions analyse human skill, labor and judgment. Copyright protection will similarly likely be refused for works generated entirely by AI. Nonetheless, the existence of humans in the creation process of the final design may grant the design some protection. This absence of clear legal guidance would mean Indian fashion designers and technology companies would be working without certainty. Without such a law, the courts would need to find author on a case by case basis.
Ownership Challenges in AI-Generated Fashion Designs
Once you start having more than one person then ownership gets really messy. AI-generated fashion designs can involve:
- The individual or entity that created the AI.
- The organization which trained the model.
- The user to whom the command was addressed.
- The brand who will achieve the peak design.
Each person is contributing in their own unique way and I don‘t really see the existing laws of copyright to be up to scratch in coping with this. Problems for policy perspective “the owner will only grant usage rights if the uses are profitable for them and for creative users will probably offer no incentives to develop new uses”. That it is wrong to let users own everything could undermine the progress of AI tools instead. A system of right should be balanced in creativity and innovation.
Adequacy of Existing Copyright Frameworks
AI creativity can not be controlled by current copyright law. The need for a human author means many AI created works will remain uncopyrighted, and the exceptions to the copyright laws were not conceived to encompass today‘s AI. In particular the fashion industry and garment industry are at risk because designs though intrinsically valuable are easy to be reproduced. Unclear ownership rights may therefore be a deterrent to firms’ adoption of AI, leading to a rise in the volume of litigation. Therefore with these factors considered more new legislation should be created relating to the authorship, ownership, and infringement of AI designed fashion. The legal system must give the legislator the freedom to provide a special copyright law for computer-generated works rather than forcing them to make use of a time-worn body of law designed for human-generated copyright works.
CASE LAW DISCUSSION
Eastern Book Company v D B. Modak (2008) 1 SCC 1
Facts
An authorship dispute occurred when Eastern Book Company (EBC), which published the Supreme Court Cases (SCC) law reports, claimed that the alterations to the Supreme Court judgments made by the published EBC versions had been plagiarized by D. B. Modak et al.
The defendants’ contention was that since decisions were public documents, they could not be copyright protected.
Judgment
The Supreme Court held that simply working hard or making a capital outlay should not be enough. It posited that creative works were the only subject matter for copyright. Although the original document the judicial decisions was free of copyright, the modifications undertaken by EBC in re-numbering the paragraphs, correcting errors, and adding footnotes involved creative works.
Legal Principle Established
The Court observed that in India the copyright stems from creation and the author must impart something of his originality. The act of doing is not enough.
Relevance to AI-Generated Fashion Designs
This case holds importance with regard to originality in rules for AI-created designs for clothing in India. In cases where a full design is made by the AI (no human involvement in the design itself) the design might not be considered original. Where the designer is exercising some degree of own judgment during the design process (through choice of prompts, modifying the outputs) and has proceeded with further design work based upon those AI outputs then the work might have copyright.
Microfibres Inc. Girdhar & Co. 2009 SCC OnLine Del 1647
Facts
It was also stated by Microfibres Inc. That some of their upholstery and fabric designs were copied by Girdhar & Co. The real problem in this case was to ascertain whether the designs in question should be protected under the Copyright Act or under the Designs Act, 2000.
Judgment
It was clarified by the Delhi High Court that Copyright cannot extend beyond the experimental phase of the mass production of the article which is registered or is registrable under the Designs Act. This protection of copyright may not, however, be perpetuated if the work falls into the classification of an industrial design.
Legal Principle Established
Having explained the relevance of the case the author considered the way in which the rights of design and copyright have been welded together. Products that are unformed at first in a mass producer are almost exclusively covered by the Designs Act.
Relevance to AI-Generated Fashion Designs
The precedent set by this case and is often termed landmark is of huge benefit to fashion. But a textile design or an illustration of a garment generated by AI not by an American AI may be said to ” be regarded as an artwork” until it is embodied in an item of fashion (mass fashion garment or product etc.) and in such a situation, copyright may not be available unless registration is applied. Therefore the application of both copyright and designs on fashion innovation may need consideration.
Thaler v. Perlmutter 2023 WL 5333236 (D. D. C. 2023)
Facts
Stephen Thaler tried to get copyright registered on a piece of artwork. “A Recent Entrance to Paradise” was designed by a machine, and the program was called the Creativity Machine. But Thaler argued that only the AI was the creator. So in every aspect it belonged to him. Since AI‘s owner was Min, so it must belong to Min.
Judgment
The District Court agreed that the copyright registration had been properly refused by the Copyright Office. The Court held that only a human author may copyright and works produced by a machine do not possess copyright protection.
Legal Principle Established
This confirmed the principle of human authors, which says that copyright law is applied for the present at computers but humans.
Relevance to AI-Generated Fashion Designs
This is one of the greatest cases in recent AI author history. The implications for fashion design is vast. An AI-created work, such as clothing design, a textile print or fashion illustration could be considered an infringement unless there is a substantial human contribution under current law. This situation also highlights the gap that already exists between new technology and existing copyright law and highlights the need for fresh legislation governing AI-created works.
Comparative Significance
All these examples highlight three significant legal challenges facing AI- generated fashion design: Eastern Book Company laid down what was required to be original, while Microfibres confirmed the difference between a design right and copyright to the context of fashion for commercial purposes, and Thaler solidified that human authorship is required. It appears safe to say, with these three cases to build upon, that if current IP laws were conceived with human authors and owners in mind, they are indeed failing. Legislation for AI fashion is certainly required.
CRITICAL ANALYSIS AND FINDINGS
As one could see from above analysis, there is a big discrepancy between the new technology and the old copyright law. The framework of those laws remain focused on humans and incorporate an understanding of creativity that pre-dates AI. The core issue is thelack of any standard definition for the authorship of AI. There is no special provision within the Indian and US Copyright Law about machine generated work. The law of computer generated work in UK was however brought before any complex AI development had happened, therefore maybe not apt to current times. Another question is about ownership. While there are many people responsible for the production of computer-generated designs for fashion, the standard under copyright law is that there should be one master author. The interaction between the developers, data suppliers, user and company are not taken into account. Different countries have different comparisons too. So, US is human-oriented. UK has a more flexible IP system compared with US, while India stays in the middle, but unclear on whose better. The variations mentioned above have a high possibility to encounter with MNC fashion firms. But on the other hand, if all copyright protection was denied it could prevent producers and copyright owners from taking the risks they take and make investments. On the other hand, to include the full protection may lead to monopolistic creativity over AI created content and can be contrary to the objectives of Copyright. It would be most advantageous for the creation of a new legislation dedicated exclusively for AI generated works. This type of law would need to address authorship, split ownership rights, create clear guidelines, as well as, be open to both human art, and the development of such.
CONCLUSION
AI technologies are changing the process of making clothes designs, producing complicated, marketable, and quickly in simple human touch. However, the law of copyright remains tied to older, more traditional ideas of human creativity and authorship and therefore unclear as to who is the “owner” and who may “prevent” an AI-created fashion design. This article explains that AI issues are only addressed partially by Indian, UK and US law systems.The decisions made by the court remain human-centered despite the fact that the law fails to acknowledge the ingenuity of the computer in its decisions. In conclusion, it would seem the central question has been answered. Current copyright laws do not take into account the difference between creation and ownership with regard to AI assisted fashion design sufficiently. Sure there are going to be laws that work for new technology but we have big reforms of the law needed. Ownership should be explicitly defined, diverse human contributions should be accounted for, and art created by AIs should be awarded its own unique protection by law in the future. Such changes would result in legal certainty, foster innovation and guarantee that copyright law progresses at the same speed as AI‘s growing relevance in the sphere of creative expression.
REFERENCE(S):
Cases
Feist Publications, Inc v Rural Telephone Service Co., 499 US 340 (1991).
Thaler v Perlmutter, 2023 WL 5333236 (DDC 2023).
University of London Press Ltd v University Tutorial Press Ltd [1916] 2 Ch 601.
Legislation
Berne Convention for the protection of literary and Artistic works 1886.
The copyright act 1957 of India.
Copyright, Designs and Patents Act 1988 (UK).
Books and Journal Articles
Andres Guadamuz, ‘Artificial Intelligence and Copyright’ (2017) 41 WIPO Magazine 14.
The new article by Jane C Ginsburg and Luke A Ali Budiardjo “Authors and Machines,” published in 2019, in the Berkeley Technology Law Journal,34 Berkeley Tech. L. J. 343.
Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law (Cambridge University Press 2020).
Susan Scafidi ‘Intellectual Property and Fashion Design’ (2007) 1 Intellectual Property and Information Wealth 115.
WIPO, WIPO Conversation on Intellectual Property and Artificial Intelligence (WIPO 2023).
Online Sources
US Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence (2023)
World Intellectual Property Organization, ‘Artificial Intelligence and Intellectual Property Policy’ (2023).
