By Tan Yen Lin Alyssa
Introduction
In the recent case of THJ Systems Ltd v Sheridan [1], the England and Wales Court of Appeal addressed issues surrounding Graphic User Interfaces (GUIs), setting a precedent for treating software-produced outputs. The case focuses on whether GUIs qualified for copyright protection under the Copyright, Designs and Patents Act 1988 (CDPA), and sought to determine the rightful owner of its copyright. This article explores the impacts of THJ v Sheridan for Computer Generated Works (CGW) and AI-generated works under UK copyright law, specifically whether these outputs meet the “originality” standard and ownership of their copyright while drawing comparisons with Singapore’s approach.
The Facts
Mitchell, a software developer, and Sheridan, a former options trader turned mentoring services provider, were both members of OptionNet LLP. Mitchell developed a software called OptionNET Explorer (ONE) to assist with trading options for his company, THJ Systems Ltd. Sheridan would use ONE to display data and information to his clients, which would in turn benefit Mitchell by promoting the use of the ONE platform. The LLP agreements required copyright notices on all materials involving the software. When Sheridan was later expelled from the LLP, Mitchell claimed that Sheridan infringed copyright laws in the software by using it after his expulsion and expiry of the licence agreement. Furthermore, both Mitchell (the developer) and Sheridan (the user) claimed to own the visual images under section 9(3) of the CDPA, which states that for “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.” The two key issues in the case relating to copyright protection will be examined below.
The Originality Standard
The first issue concerns whether copyright subsists in certain GUIs in the form of visual risk and price (R&P) charts produced using the ONE software. GUIs act as “the interface which enables communication between the software and user” [2]. Following section 4(1)(a) of the CDPA, which defines an artistic work as “a graphic work, photograph, sculpture or collage, irrespective of artistic quality,” the claimants contend that the R&P charts qualified as artistic works, specifically graphic works. For copyright to subsist in an artistic work, it must be “original” as stated in section 1(1)(a) of the CDPA [3].
Traditionally, English copyright law had adopted the “skill, labour and judgment” test when approaching the issue of whether copyright subsists in a work. In the initial judgement, the High Court applied the established “skill and labour” test, ruling that Mitchell’s contributions in shaping the GUIs’ “look and feel” through fonts, layouts and colours constituted sufficient “skill and labour” to qualify as artistic works. This is similar to the GUIs in Navitaire Inc v Easyjet [4], where Pumfrey J noted that “to arrange a screen certainly affords the opportunity for the exercise of sufficient skill and labour for the result to amount to an artistic work” and concluded that the GUI screens in that case satisfied this requirement [5].
However, following the landmark decision in Infopaq International A/S v Danske Dagblades Forening [6], the Court of Justice of the European Union (CJEU) reaffirmed that the appropriate standard is the “author’s own intellectual creation” test [7]. On appeal, Arnold LJ highlighted that the judge erred in relying on the obsolete test of “skill and labour”. Instead, the more rigorous test of the “author’s own intellectual creation” should have been applied. Reinforced by Painer v Standard Verlags GmbH [8], Arnold LJ argued that originality hinges on the author’s ability to “express their creative abilities… by making free and creative choices so as to stamp the work created with their personal touch [9].” Since there were “choices as to what to put where, including such matters as which commands to put into the ribbon and in what order" for the R&P charts, the court found that it met the originality standard [10]. Furthermore, in accordance with the CJEU’s interpretation, Arnold LJ reaffirmed that “the test for originality is an objective one” without consideration of artistic merit, which accords with the statutory wording in Section 4(1)(a) of the CDPA [11].
However, the originality test should be revisited to better accommodate its application to CGW and AI-generated works. Initially, it may seem that CGW satisfies the “author’s own intellectual creation” test established in THJ v Sheridan. If the author had meaningful control over the generative process, such as making deliberate design choices and curating prompts, the work can be said to have been created in a way that allows the author to “make free and creative choices” with an element of their “personal touch”. For example, The US Copyright Office recognises that works containing AI-generated material can qualify for copyright if they “contain sufficient human authorship.” [12] This is provided the author maintains control over the traditional elements of authorship, by selecting or arranging AI-generated material in a “sufficiently creative way that the resulting work as a whole constitutes an original work of authorship [13].” If we apply this to the more stringent UK originality standard, input prompts could qualify as “literary works” if sufficiently original. However, the outputs may not necessarily bear the user’s “personal touch” because users do not exercise sufficient control over the training data and operation. Even with detailed prompts, the user’s influence on the final output may be too limited to meet the originality standard [14]. Thus, it would be difficult for UK courts to find that a user exercised the level of creative decision-making seen in cases like Painer, where originality was established through choices of “the framing, the angle of view and the atmosphere created”.
Upon closer inspection, this test poses challenges to CGW in cases where outputs are fully automated. Arnold LJ clarified that works “dictated by technical considerations, rules or other constraints which leave no room for creative freedom” [15] fail to meet the standards of the test. For instance, in SAS Institute Inc v World Programming Ltd [16], the European Court of Justice held that CGW produced by software, such as languages and data visualisations, lacked originality because they were generated based on technical rules and parameters embedded in the software itself. The court emphasised that, in line with the principle that “only the expression of a computer program is protected by copyright,” logic, algorithms and programming languages are not protected [17], and that “neither the functionality of a computer program nor… the format of data files” constitutes copyrightable expression [18]. Hence, CGW generated by fully automated processes,and constrained by its algorithms or technical parameters are unlikely to meet the test.
By implying that CGW and AI-generated works must meet the same rigorous originality standards as human-authored creations, the current provisions do not seem to adequately address the unique nature of these outputs. CGW often results from complex, automated processes driven by algorithms or machine learning models, rather than human creativity alone. While a human author can make free, deliberate choices that influence every aspect of their work, it is impossible to control the exact outcome in CGW as these systems generate multiple variations of outputs based on programmed parameters [19]. As a result, these works may not reflect the “personal touch”, making it challenging to meet these standards of originality. Moreover, the lack of copyright protection for AI-generated output would leave creators who use generative AI tools in their work vulnerable. Without adequate protection for their AI-generated or AI-assisted works, their works would fall into the public domain, allowing anyone to freely access and reuse these works without infringing on copyright [20]. This stifles creativity, as creators are discouraged from using generative AI tools knowing that their contributions cannot be protected against infringement or unauthorized use [21]. Hence, there is a compelling need for a reevaluation of the originality test, particularly for works created with and by AI.
Ownership of Copyright
Secondly, the issue of whether the developer or user of the software is the true owner of its output raises important concerns regarding the application of Section 9(3) of the CDPA for CGW and AI-generated outputs.
Under Section 9(3) of the CDPA, for computer-generated works, “the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken”. In the case, Mitchell argued that, as the developer, he should be considered the author of the R&P charts generated by the ONE software, since he made the “necessary arrangements” for the software to produce these images [22]. He designed the software to generate the R&P charts by providing the technical framework necessary for their creation. Conversely, the defendants argued that the user should be the author instead of the developer, as it was the user's actions — the selecting of data and configuring of the time period for the charts—that were integral to the creation of the images. The user is the person who made the arrangements necessary for the creation of the work, fulfilling the requirement of Section 9(3) and thus claiming ownership of the copyright in the images [23]. Authorship was not directly addressed on appeal, but the Court of Appeal appeared to accept Mitchell as the author, focusing on his intellectual input in designing the user interface, with Sheridan’s operation of the software deemed irrelevant.
The case brings up a significant concern in the application of Section 9(3) to CGW and AI-generated works. There is an inherent contradiction in the definition of CGW as a work “generated by computer in circumstances such that there is no human author of the work" and the originality test above, which fundamentally relies on the “author’s own intellectual creation”. Without a human author to demonstrate “intellectual creation”, CGW and AI-generated outputs will not satisfy the originality test in the first place. If the originality test takes precedence, this means that Section 9(3) becomes effectively redundant. Additionally, there is ambiguity over who qualifies as the “person by whom the arrangements necessary for the creation of the work are undertaken.” Is it the developer who designed the AI programme and was responsible for training it, or the user who created its resulting output through specific prompts? If the originality test takes precedence and is applied, the user of the generative AI model would qualify as the author as long as their prompts involve “free and creative choices” that satisfies the requirement of “author’s own intellectual creation”. However, academics like Guadamuz argue that Section 9(3) operates as an exception to the originality standard, as the law contends there is a lack of creative input in CGW [24]. He cites Lord Beaverbrook during the 1988 House of Lords debates on CGW, that the individual making the arrangements necessary for such creations “will not himself have made any personal, creative effort [25]”. This view aligns with the decision in THJ v Sheridan, implying that AI-generated outputs would have the developer who is responsible for training or organising the AI's development as their author [26].
The question of who rightfully owns the output of CGW and AI tools remains a largely contentious and unresolved issue. Despite THJ v Sheridan seeming to favour the developer as the author, in a public discussion of the case, Arnold LJ declined to take a clear stance on whether generative AI outputs qualify for under UK copyright protection, acknowledging that this issue is “likely to be the subject of future litigation”. While critics have proposed alternative solutions, such as redefining “originality” to focus on the “novelty” of the work [27], neither the judiciary nor the legislature has embraced these approaches. This lack of clarity leaves creators using AI tools uncertain about their rights, hence addressing this would ensure copyright law remains sufficient to meet the modern demands of AI.
Comparative Perspective: Singapore’s Test
Singapore similarly requires work to be "original" for copyright to subsist [28]. Similar to the UK’s “author’s own intellectual creation” test, CGW may qualify as "literary works" under subsection (3) of the Copyright Act 2021 if they demonstrate “intellectual creation because of the selection or arrangement of their contents” [29].
However, Singapore adopts a stricter test to determine authorship for CGW as compared to the UK. In contrast to the UK’s broad “author’s own intellectual creation” test, Singapore focuses on identifying the individual who exercised control over the “creative aspects” of the work. This places a greater burden on claimants to have evidence of direct, creative involvement as compared to simply general supervision or preparatory arrangements. In Global Yellow Pages Ltd v Promedia Directories Pte Ltd [30], which involved a compilation of data, the Singapore Court of Appeal held that only efforts directly related to creation are considered relevant, such as selecting and arranging the data. This means that “efforts directed toward… gathering of training data” into computer or AI systems are not relevant [31], which makes the user more likely to be recognised as the author due to their direct influence over the output. Additionally, in Asia Pacific Publishing Pte Ltd v Pioneers & Leaders (Publishers) Pte Ltd [32], the Singapore Court of Appeal held that since there is no “guiding principle as to the precise quantum of labour, skill or judgment required”, copyright for human authors in Singapore must be “determined on the facts of the particular case [33]”. This means the Singapore courts assess the degree of control the author has over the creative aspects of the work, which makes the user of the generative AI tool– the one who selects, arranges, or modifies the content– more likely to be considered the author.
Nevertheless, Singapore currently does not have any copyright provisions regarding AI-generated outputs [34], while the UK provides explicit copyright protection for CGW without a human author under Section 12(7) of the CDPA for 50 years from the date they are created [35]. This allows the possibility of AI-generated outputs qualifying for copyright protection, even in cases where no specific human author is identifiable [36]. The difference between Singapore’s more restrictive framework and the UK’s more flexible approach highlights the two countries’ varying priorities in dealing with copyright for CGW and AI-generated content. While Singapore focuses on maintaining its rigid standards, the UK is shifting more towards accommodating AI-assisted outputs within its copyright provisions. However, if Singapore adopts Whitford J’s perspective in Express Newspapers v Liverpool Daily Post [37], which views computers as “no more than the tool” used by the human author to create works, there could be potential for recognising AI as a “tool" for creating the end product [38]. Since it would be unrealistic to attribute authorship to the AI system rather than the user controlling it [39], AI-generated outputs could qualify for copyright protection with sufficient human authorship [40].
Conclusion
Despite setting a promising precedent for CGW and AI-generated works under UK copyright law, THJ v Sheridan emphasises the need to address the underlying issues and contradictions within the existing framework. Both Singapore and the UK face challenges in defining authorship and originality in the age of AI, with each country taking differing approaches. While the UK is moving toward accommodating AI-generated works, Singapore maintains its stricter standards for originality instead. Ultimately, there is a need for a more cohesive approach to copyright protection—one that not only encourages innovation but also ensures creators are sufficiently and properly protected.
Bibliography
[1] THJ Systems v Sheridan [2023] EWCA Civ 1354, [2024] E.C.D.R. 4, CA.
[2] ibid, [208]
[3] ibid, [13]
[4] Navitaire Inc v Easyjet Airline Company and Bulletproof Technologies Inc [2004] EWHC 1725
[5] ibid, [98]
[6] Infopaq International A/S v Danske Dagblades Forening [2009] ECR I-6569
[7] ibid, [35]
[8] Painer v Standard Verlags GmbH [2011] ECR I-12533, [89]-[93]
[9] ibid (n. 1), [26]
[10] ibid (n. 1), [26]
[11] Ibid (n. 1), [38]
[12] ‘Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence’ (16 March 2023) <https://www.copyright.gov/ai/ai_policy_guidance.pdf> Accessed 29 November 2024.
[13] ibid.
[14] Francesca Mazzi, ‘Authorship in artificial intelligence-generated works: Exploring originality in text prompts and artificial intelligence outputs through philosophical foundations of copyright and collage protection’, The Journal of World Intellectual Property (Volume 27, Issue 3), 26 May 2024, p. 410-427
[15] ibid (n. 1), [26]
[16] SAS Institute Inc v World Programming Ltd [2013] EWCA Civ 1482
[17] ibid, [14]
[18] ibid, [46]
[19] Hafiz Gaffar and Saleh Albarashdi, ‘Copyright Protection for AI-Generated Works: Exploring Originality and Ownership in a Digital Landscape’, Cambridge University Press (23 January 2024)
[20] ibid.
[21] Jie Zhang and Xin Xie, ‘Research on the Copyright Protection of Artificial Intelligence Generation in the Smart Media Environment’, 2023 3rd International Conference on Public Management and Intelligent Society (PMIS 2023), 665
[22] ibid (n. 1), [38]
[23] ibid (n. 1), [49]
[24] Andre Guadamuz, ‘Do Androids Dream of Electric Copyright? Comparative Analysis of Originality in Artificial Intelligence Generated Works’, Oxford University Press (2021)
[25] HL Deb vol. 493 col. 1305. 25 February 1988
[26] ibid (n. 1), [216]
[27] Lionel Bently, Brad Sherman, Dev Gangjee & Phillip Johnson, Intellectual Property Law 5th ed (OUP, 2018), 117
[28] Copyright Act 2021, s 109(1)(a)
[29] ibid. Subdivision (3), s 13(1)-(2)
[30] Global Yellow Pages Ltd v Promedia Directories Pte Ltd [2017] SGCA 28
[31] Pin-Ping Oh, ‘Copyright protection for AI-generated works in Singapore’ (Bird & Bird, 1 August 2022) <https://www.twobirds.com/en/insights/2022/singapore/copyright-protection-for-ai-generated-works-in-singapore> Accessed 29 November 2024
[32] Asia Pacific Publishing Pte Ltd v Pioneers & Leaders (Publishers) Pte Ltd [2011] SGCA 37
[33] ibid, [34]
[34] ibid (n. 27), s 115
[35] Copyright, Designs and Patents Act 1988, s 12(7)
[36] ibid (n. 30)
[37] Express Newspapers Plc v Liverpool Daily Post & Echo Plc [1985] 1 WLR 1089
[38] ibid (n. 30)
[39] ibid (n. 36), [1098] (Whitford J)
[40] ibid (n. 30)
Comments