AI Art and the Limits of Intellectual Property Law
Guest Author: Periklis Agalopoulos
Editors: Laureen Wong and Rishabh Chawda
Introduction
This paper examines whether art generated by artificial intelligence (AI), can fall within the ambit of the Copyright Designs and Patents Act 1988. Specifically, it assesses the potential of AI generated art as a legitimate candidate to gain authorship or/and ownership of the produced piece of artefact. This essay initially seeks to elaborate on the definition/nature of modern AI and examines the UK’s current legal framework on copyright whilst critically comparing it to the European Union’s (EU) relevant regime. It sets out the roles and possible legal rights that different entities such as programmers, users, curators of datasets, and AI trainers have in the process of an AI-generated artwork. Lastly, this essay proposes methodologies under which a clear delineation can be made as to the rights that each individual, or the AI, may have under copyright law.
It is imperative to examine whether artworks generated by AI are eligible to be attributed copyright protection within the area of Intellectual Property (IP) law. In doing so, this paper scrutinises the interchangeable relationship of the different entities that participate in the ‘creation’ of the work and could therefore claim potential ownership of it. Additionally, this essay analyses the extent to which the current international legal frameworks provide such protection.
Definition of AI
The concept of AI is highly convoluted[1] because academics[2] in this field have proposed a plethora of interpretations[3], ultimately failing to validate a uniformly recognised definition. The first attempt at a coherent definition can be traced back to 1946, when Alan Turing stated that AI was the idea of a computer showing ‘intelligence’[4], with chess-playing as a paradigm. Various other theories build upon this and describe it as being an “activity devoted to making machines intelligent; and intelligence is that quality that enables an entity to function appropriately and with foresight in its environment”[5]. It is evident that AI has a variety of ‘faces’ and characteristics which are capable of ‘mimicking human behaviour’[6]. Despite the fact that they can be commonly referred to as ‘machines’,[7] most modern-day AI function together with Machine Learning (ML)[8]. ML is a branch of AI which ‘focuses on developing systems that learn from the assimilation of data’[9] and allow AI to ‘improve overtime on a specific task’[10]. ML is an algorithmic technique created by programmers that significantly increases the capabilities of AI to ‘learn’ to generate outputs based on the input provided by humans, without being ‘pre-programmed’ to produce those specific results.
This extension of AI has raised controversy[11] as this type of training has incrementally minimised human intervention with respect to the production and execution of computer-generated /AI-assisted works. This can also be illustrated by the development of Deep Learning and Generative Adversarial Network (GAN), a subdivision of ML which has “transformed the way in which algorithms achieve (or exceed) human-level performance in areas such as game playing and computer vision”[12]. Arguably, it further restricts the participation of the human factor.
A prime example of the application of GAN is the ‘The Next Rembrandt Project’, where paintings of the Dutch artist, Rembrandt Harmenszoon van Rijn, were digitalised and used in a form of input of data that were processed by the AI through digital ‘neural networks’, in order to create a painting that emulates the painting style of the ‘authentic’ human author[13]. The final product was a unique artistic expression derived by Al’s processing/mechanical capability to deviate from the input data and form a new piece of artistic work.
Therefore, in certain circumstances, AI technology has advanced to a level that, human intervention in a work is sufficient to render it legally protectable. In the former scenario, the completed work is referred to as AI-assisted work. The AI is a mere tool[14] that formulates, in practice, a concept that was primarily conceived by human. Conversely, in the latter, the works generated are characterised as AI or computer-generated works[15], a name that reflects the minimal and sometimes missing human involvement in the execution of the project. This distinction was initially established in Express Newspapers[16], where Whitford J stated that “If you write your work with a pen, it is the pen which is the author of the work rather than the person who drives the pen”[17].
UK Copyright Law Framework
Under UK jurisdiction, copyright is regulated by the Copyright Designs and Patents Act 1988 (CDPA 1988) which contains an exhaustive list of categories[18], namely reported as subject matters where copyright protection is offered towards various intellectual properties. A piece of art would fall within the division of an artistic work[19], only if a number of criteria are met. The work in question must be ‘original’(s1(1)(a)), ‘fixated’ and ‘creative’. Fixation is not imposed by statute in artistic works, but it has been implicitly established by case law. In Merchandising Corp[20], it was held that ‘permanency’ could only be granted to a work that is placed in a legally qualified ‘surface’. The concept of originality is directly embedded within the relationship that exists between the creator and the work. The work in this context can be interpreted as “originating from the author of the work and not being a copy of a preceding work”.[21] In the case of Ladbroke,[22] the judiciary expanded this by examining whether the author of the work has exercised the requisite “labour, skill or judgment” in producing the work.
EU Copyright Law Framework
However, this test has been mostly substituted[23] by the ‘author’s own intellectual creation’[24] doctrine, which was first implemented in the European Union[25] for computer-generated works. Subsequently, it was harmonized through the decision in Infopaq[26] which extended its application to all subject matters. Further clarification of this principle was showcased by Painer[27] and Football Dataco.[28] The cases established that ‘an intellectual creation must reflect the author’s personal touch’, despite that “creative and free choices have taken place during the creation of the work”. Moreover, the Court of Justice of the European Union (CJEU) has highlighted in SAS Institute,[29] that a creator/author[30] must not be constrained by rigid rules or ‘functional limitations’[31] that deprive unique expression and creative choice, as it detracts from the credential of ‘originality’. Additionally, it can be said that the current EU legal framework has an ‘anthropocentric approach’[32] in determining originality. Therefore, the presumption of AI authorship is rebutted, because, in its current state, AI can only emulate, but not express human creativity.
AI Authorship under EU approach
As Ana Ramalho argues,[33] “the limitation of current AI programming does not allow them to make such choices”, rendering them ineligible for exclusive authorship of work. Consequently, the extension of the work-for-hire doctrine proposed by Shlomit Yanisky-Ravid[34] is inapplicable as an AI does not acquire legal personhood. Therefore, it would be contrary to public policy to acknowledge the AI as an employee. A similar approach is implicitly adopted by both German[35] and Australian[36] jurisdictions, while the United States has established an ‘empowered’ threshold for originality[37] where the court reiterated that “100 uncopyrightable facts do not magically change their status when gathered together in one place”. It thus pointed to a possible exclusion of authorship by a computer/AI.[38]
AI Authorship under UK approach
In the United Kingdom, the author of a computer-generated work is to be “the person by whom the arrangements necessary for the creation of the work are undertaken”. [39] In Nova Productions[40], Kitchin J applied this section in order to classify the programmer as the author of the frames of a computer-generated work, while rejecting the notion of the player/user being an author due to the lack of sufficient labour and artistic input into the work. This terminology, combined with s.178 of the CDPA that states that the work is generated by computer in circumstances such that there is no human author of the work, has raised issues as to the coherence of the law and the future allocation of authorship/ownership. It thus, ultimately, circumvents the requirement for human authorship and originality imposed by statute in ‘normally’ produced artistic works.
This stance seems to favour AI authorship and contradicts the EU approach because it suggests that these types of works do not fall directly within a human creator; signified also by s12(7) of the CDPA, which bestows lesser copyright protection compared to other types of works. On this point, Benjamin Williams said, “we are in a strange position where the section applies to works which putatively have no human author, but which are only capable of protection if they are the original creation of an author.”[41]
AI or Human? Distinguishing the Creator
In light of the aforementioned issue, the most suitable method to distinguish who is/are the human author(s) of this type of work, is to utilise the approach taken by the CJEU in Painer and proposed by the EU Commission,[42] which purportedly segregated the creative process into ‘conception’, ‘execution’ and ‘redaction’. The first stage pertains to the pre-production phase, where programmers may create the code for the creation of the AI. Developers then conceptualise, in an elucidated manner, the theme and curators will concentrate the data that will be used in the project. This primal ‘step’ has been integrated into the ‘creative effort and originality test’ and, therefore, increases the probability for these entities to be considered as creators/authors of the work.
The main hurdle which blurs and undermines human creativity is the ‘execution’ stage, where the Al may fulfil the processing of data and any relevant action independently, without any human actor. This conduct, adjoined by the ‘black box’ phenomenon[43], where the people involved are incapable of preconceiving the exact output, strains not only the originality but also the ‘humane’ creative expression that must be present in the computer-generated artwork. However, as Hugenholtz points out,[44] this can be reconciled by the ‘redaction’ stage in which humans would invariably make final amendments to the output, with the aim of rendering it commercially exploitable. This is an essential role which provides the opportunity for people to display control over the work and directly express a form of ‘personal touch’ in the final product. It promotes their eligibility to be acknowledged as creators/authors or joint-authors.[45]
In particular, as Alina Skiljic suggests,[46] a case-by-case approach is a justified manner under which authorship can be efficiently scrutinised. Thus, when a programmer directs a parameter that guides the AI, and that he/she is solely responsible for undertaking the primary creative work of the project, he/she should be considered the author/owner[47] of the AI as well as the owner of its derivative work. On the other hand, exclusive authorship on one category of creators seems implausible, especially in computer-generated artwork. This is because, in principle, there is a variety of human contributors for a piece of artefact to be construed.[48] As a result, in these types of settings, multiple human actors can acquire joint authorship and ownership of the work.
This type of authorship arises only if it can be established that the “work produced by the collaboration of two or more authors” is such that “the contribution of each author is not distinct from that of the other author or authors”(s10(1) CDPA). On this note, the Martin[49] case illustrates that the UK courts will adopt a contextual approach in evaluating the level of contribution, cooperation and mutual coordination/common purpose between the entities. It is important to identify who among them had the ‘final say’ on what will be included.[50] The test for reaching the threshold for joint authorship will be ascertained on a factual basis determined by the satisfaction of these requirements in the various stages of the project. Consequently, if the user of the AI only ‘pushed a button’[51] when creating an artwork, this action will not satisfy the criterion of ‘creativity’. If the user or any other entity does not coordinate with the other creators and just finds the AI ‘ready’ for independently creation of work, it will also fail for the test of ‘creativity’.[52] In this example, the user could still receive ownership of the Al’s output through the licencing[53] of the AI by the owner which provides potential right of exploitation in exchange of monetary compensation. By contrast, if the user or any other moderator (such as the curators of the dataset or the AI trainers of the work) participated vigorously in the redaction stage by transforming the output into a commercially viable product, they could potentially be considered owners of the work.
In parallel, the promotion of joint authorship does not only prevent this type of work from reaching the public domain, ‘nullifying’ human creativity,[54] but it can also combat the issue of prospective monopolisation[55] with respect to programmers or companies. It provides the necessary economic incentives for users and other experts to innovate and participate in technological evolution and human welfare. Lastly, problems deriving from multiple ownership of an artefact can be remedied by contractual clauses, as Mark Hyland emphasises,[56] “either by making the collaborator an employee and thus surrendering their claim, or alternatively, by insisting on the grant of an option for the assignment of any rights accrued for a pre-determined fee”.
Conclusion
In conclusion, the current legal framework does not support AI authorship for computer-generated work. Although the outcome of ownership is preserved among humans, this situation itself is plagued with discrepancies. An amicable solution is for the courts to separate each part of the process of the work according to originality/creativity, thereby recognising authorship and the right of ownership. Despite the courts in various jurisdictions having not yet faced the dilemma arising from AI ownership, it can be argued that in the foreseeable future, AI shall satisfy the ‘cogito ergo sum’[57] theory. Consequently, the law would be bound to ameliorate the present regime with the aim of accommodating the gaps of intellectual property law.
[1] Bram Van Wiele, ‘The Human-Machine Synergy: Boundaries of Human Authorship in AI-Assisted Creations’ (2021) 43 EIPR 164, 166.
[2] Katarina Foss-Solbrekk, ‘Three Routes to Protecting AI Systems and their Algorithms Under IP Law: The Good, the Bad and the Ugly’ (2021) 16 JIPLP 247, 248.
[3] Josef Drexl, ‘Technical Aspects of Artificial Intelligence: An Understanding from an Intellectual Property Law Perspective’ (2019) Max Planck Institute for Innovation and Competition Research Paper No 19-13, 1 <https://ssrn.com/abstract=3465577> accessed 8 February 2022.
[4] Alan M Turing, ‘Computing Machinery and Intelligence’ (1950) 59 Mind New Series 433.
[5] Nils J Nilsson, The Quest for Artificial Intelligence: A History of Ideas and Achievements (Cambridge University Press 2010) 1.
[6] Josef Drexl, ‘Technical Aspects of Artificial Intelligence: An Understanding from an Intellectual Property Law Perspective’ (2019) Max Planck Institute for Innovation and Competition Research Paper No 19-13, 3 <https://ssrn.com/abstract=3465577> accessed 8 February 2022.
[7] Timothy L Butler, 'Can a Computer be an Author - Copyright Aspects of Artificial Intelligence' (1981) 4 Hastings Comm & Ent LJ 707, 708.
[8] Zack Naqvi, ‘Artificial Intelligence, Copyright, and Copyright Infringement’ (2020) 24 Marq Intell Prop L Rev 15, 18.
[9] European Commission, ‘Trends and Developments in Artificial Intelligence Challenges to the Intellectual Property Rights Framework’ Final Report 1, 25.
[10] Harry Surden, ‘Machine Learning and the Law’ (2014) 89 Wash L Rev 87, 91.
[11] Bram Van Wiele, ‘The Human-Machine Synergy: Boundaries of Human Authorship in AI-Assisted Creations’ (2021) 43 EIPR 164, 165.
[12] European Commission, ‘Trends and Developments in Artificial Intelligence Challenges to the Intellectual Property Rights Framework’ Final Report 1, 28.
[13] Shlomit Yanisky-Ravid, ‘Generating Rembrandt: Artificial Intelligence, Copyright, and Accountability in the 3A Era: The Human-like Authors Are Already Here: A New Model' (2017) 2017 Mich St L Rev 659.
[14] Benjamin Williams, ‘Painting by Numbers: Copyright Protection and AI-Generated Art’ (2021) 43 EIPR 786, 788.
[15] Emily Dorotheou, ‘Reap the Benefits and Avoid the Legal Uncertainty: Who Owns the Creations of Artificial Intelligence?’ (2015) 21 CTLR 85.
[16] Express Newspapers v Liverpool Daily Post & Echo [1985] 1 WLR 1089 (HC).
[17] ibid [1093].
[18] Copyright Designs and Patents Act 1988, s1.
[19] Copyright Designs and Patents Act 1988, s4.
[20] Merchandising Corp v Harpbond [1983] 3 FSR 32 (CA).
[21] University of London Press v University Tutorial Press [1916] 2 Ch 601, 609.
[22] Ladbroke v William Hill [1964] 1 All ER 465, 475.
[23] Andres Guadamuz, ‘Do Androids Dream of Electric Copyright? Comparative Analysis of Originality in Artificial Intelligence Generated Works?’ (2017) 2 IPQ 169, 177.
[24] Council Directive 96/9/EC of 11 March 1996 on the legal protection of databases Article 3(1) OJ L 77/20.
[25] Council Directive 2009/24/EC of 23 April 2009 on the legal protection of computer programs OJ L 111/16.
[26] Case C-5/08 Infopaq International A/S v Danske Dagblades Forening [2009] ECR I-06569.
[27] Case C-145/10 Eva-Maria Painer v Standard VerlagsGmbH and Others [2011] ECR I-00000, paras 87-93.
[28] Case C-604/10 Football Dataco Ltd and Others v Yahoo! UK Ltd and Others [2012] 2 CMLR (24) 703 (ECJ).
[29] Case C-406/10 SAS Institute v World Programming [2012] 3 CMLR (4) 259 (ECJ).
[30] Copyright Designs and Patents Act 1988, s9.
[31] Case C-393/09 Bezpečnostní softwarová asociace v Ministerstvo kultury [2010] ECR I -13971.
[32] Enrico Bonadio and Luke McDonagh, ‘Artificial Intelligence as Producer and Consumer of Copyright Works: Evaluating the Consequences of Algorithmic Creativity’ (2020) 2 IPQ 112, 115.
[33] Ana Ramalho, ‘Will Robots Rule the (Artistic) World? A Proposed Model for the Legal Status of Creations by Artificial Intelligence Systems’ (2017) J Internet L 1, 8 (forthcoming).
[34] Shlomit Yanisky-Ravid, ‘Generating Rembrandt: Artificial Intelligence, Copyright, and Accountability in the 3A Era: The Human-like Authors Are Already Here: A New Model' (2017) 2017 Mich St L Rev 659, 707.
[35] German Act on Copyright and Related Rights, Article 11.
[36] Australia Copyright Act 1968, s10.
[37] Feist Publications Inc v Rural Telephone Service Co 111 S Ct 1282 (1991).
[38] Compendium of US Copyright Office Practices, s306.
[39] Copyright, Designs and Patents Act 1988, s9(3).
[40] Nova Productions Ltd v Mazooma Games Ltd [2007] EWCA Civ 219, [2007] EMLR 14.
[41] Benjamin Williams, ‘Painting by Numbers: Copyright Protection and AI-Generated Art’ (2021) 43 EIPR 786, 790.
[42] European Commission, ‘Trends and Developments in Artificial Intelligence Challenges to the Intellectual Property Rights Framework’ Final Report.
[43] Josef Drexl, ‘Technical Aspects of Artificial Intelligence: An Understanding from an Intellectual Property Law Perspective’ (2019) Max Planck Institute for Innovation and Competition Research Paper No 19-13, 11 <https://ssrn.com/abstract=3465577> accessed 8 February 2022.
[44] P Bernt Hugenholtz and Joao Pedro Quintais, ‘Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output?’ (2021) 52 IIC 1190, 1203.
[45] Copyright, Designs and Patents Act 1988, s10(1).
[46] Alina Skiljic, ‘When Art Meets Technology or Vice Versa: Key Challenges at the Crossroads of AI Generated Artworks and Copyright Law’ (2021) 52 IIC 1338, 1349.
[47] Copyright, Designs and Patents Act 1988, s11(1).
[48] Nina I Brown, ‘Artificial Authors: A Case for Copyright in Computer-Generated Works’
(2018) 20 Colum Sci & Tech L Rev 1, 4.
[49] Martin v Kogan [2019] EWCA Civ 1645, [2020] EMLR 4.
[50] ibid [59].
[51] Grimmelmann, ‘There’s No Such Thing as a Computer-Generated Work’ (2016) 39 Columbia Journal of Law & the Arts 404, 411.
[52] Martin Senftleben and Laurens Buijtelaar, ‘Robot Creativity: An Incentive-Based Neighbouring Rights Approach’ (2020) 42(12) EIPR 797, 807.
[53] Copyright, Designs and Patents Act 1988, s90.
[54] Wietse Vanpoucke, ‘Copyright Challenged by Art Created by Artificial Intelligence’ (2021) 43(8) EIPR 495, 501.
[55] Sarah Legner, ‘Are Works of Artificial Intelligence in Need for Further Protection’ (2021) 43(2) EIPR 71, 73.
[56] Mark Hyland and Michael Howard, ‘Case Comment: Martin v Kogan [2017] EWHC 2927’ (2018) 40(12) EIPR 817, 820.
[57] René Descartes, ‘Discourse on Method’ (1637).
