Unraveling the Copyright Conundrum: AI-Generations in the Legal Spotlight
In the realm of copyright, the state of law remains largely convoluted due to conflicting precedent and incongrous discussions among academics and creative industries. The seemingly practical yet alluring invention of AI systems like ChatGPT, Bard, DALL-E, Midjourney and Stable Diffusion are not only devouring the production of literary, dramatic, musical and artistic (“LDMA”) works, but also projecting a disenchanting fact that humans are no longer indispensable in the creative process.[1] Hence, this article aims to first enunciate the current law on copyright, followed with an exploration into the notions of originality and authorship in AI-generated works, and lastly a brief evaluation of the normative justifications for the copyright protection of AI-generated works.
Examples of AI-Generated Works
AI technologies have massively been harnessed in creative industries to generate LDMA[2] outputs that would normally be protected under copyright law.[3] For instance, in relation to artistic works, the first AI-generated artwork “Comte de Bellamy” reproduced the same features and style of Edmund Belamy, which was subsequently sold at Christie’s auction for $432,500.[4] Another example is the portrait entitled “The Next Rembrandt”, which was generated by computer upon analysing thousands of art pieces created by the 17th-century Dutch artist Rembrandt Harmenszoon van Rijn.[5] As regards to literary works, a Japanese AI program in 2016 co-authored a short-form novel titled “The Day A Computer Writes A Novel” which entered the second-round screening for a national literary prize.[6] Regarding dramatic works, AI has been exploited to duplicate actors’ facial expressions, voices and unique personal characteristics. The most recent and perhaps the most influential example of the AI intrusion on dramatic works is the dehumanizing use of AI to ‘scan background performers, have them paid for the day and then turn them into digital replicas for future use for the rest of the eternity’[7] which induced Hollywood’s longest strike organized by the Screen Actors Guild-Amercian Federation of Television and Radio Artists. Although strikes have recently came to an end, the encroachment of AI on actors and voice artists are deemed to persist with the expeditious advancement of AI. Even musical creations are not immune to AI exploitations, Deep Mind, a Google-owned AI company has created software that can generate music upon listening to recordings.[8] More recently, an AI-generated Drake and The Weeknd song – “Heart on My Sleeve” was produced utilising an AI voice-cloning software trained on musician’s voices.[9] However, the artist Drake has expressed utter displeasure on his voice being cloned – posting a 6-word statement “This is the final straw AI” on his Instagram page.[10] Nonetheless, the chart-topping French DJ David Guetta remarkably promulgated that “the future of music is in AI” upon using AI to generate lyrics and rap in the style of the US star Eminem for a live show.[11]
This controversy undoubtedly ratcheted up the public chasm between a progressive and an orthodox view of copyright protection, which boils down to two questions – whether AI-generated works deserve copyright protection or whether copyright protection should only be afforded to works created by humans.
The first AI-generated artwork “Le Comte de Bellamy”. Currently exhibited in the Museum of Civilization in Quebec, Canada. (Photo credit: Nicolas Laugero-Lassere, Obvious-Art, accessed on https://obvious-art.com/portfolio/le-comte-de-belamy/)
The portrait of Rembrandt van Rijn (15 July 1606 – 4 October 1669), who was once regarded as one of the world’s greatest painters and a key Dutch figure in Dutch representing the Golden Age of painting in the Netherlands. (Photo credit: Microsoft Corporation, accessed on https://news.microsoft.com/europe/features/next-rembrandt/)
Current Law and Further Discussion
Currently, the Copyright, Designs and Patents Act 1988 (“CDPA”) protects both authorial and entrepreneurial works. While authorial works include LDMA works,[12] entrepreneurial works encompass sound/film recording,[13] broadcasts,[14] and typographical arrangements.[15] However, only authorial works are subject to the requirement of originality,[16] which requires the work to reflect “the author’s own intellection creation” exercised through the author’s “free and creative choices” with a “personal touch”.[17] This suggests that a prerequisite to the originality requirement is that the work is exerted by a human author.[18] Further, regarding authorship, an author of an LDMA is the person who creates the work,[19] but the Act specifies that an author of a computer-generated work is deemed to be the person “by whom the arrangements necessary for the creation of the work are undertaken”.[20] This may include persons who operate the computer, provides inputs for the computer system or the programmer of the system.[21] While this definition may be conclusive to the issue of authorship in AI-generated works, the kernel of the argument is about where the originality element lies and who the ultimate author is in AI-generated works remain specious.
a) The Hunt for ‘Originality’ in AI-generated works
Given that AI falls into the category of computer-generated work, the subsistence of copyright protection in AI-generated works depends on whether the work reflects intellectual creativity exercised by the author. Although the mechanical generation of automated results does not seem to reflect any intellectual creativity by the author, it can be argued that the author “by whom the arrangements necessary for the creation of the work are undertaken”[22] has excercised “creative choices” in selecting the input data and the content used to train AI. This is exemplified in the short-form novel produced by the Japanese AI program (mentioned above) in which sentences were deliberately chosen with parameters being carefully set prior to embarking upon the construction process,[23] rendering the AI not completely autonomous in this sense. This reinforces the undisputable fact that critical and creative choices can only be exercised by humans, not AI which “cannot think critically given its lack of the ability to value”.[24]
However, the degree of originality also bears relevance in this context as it significantly influences the scope of copyright protection afforded to the work. As vividly illustrated in the landmark case of Ladbroke (Football) v. William Hill (Football),[25] a part must be “substantial” for it to be protected by copyright which is ultimately a matter of subjective impression and quality rather than quantity.[26] Accordingly, if substantial resemblance can be found between the content in the output produced by the AI and the content which is inputted by a human to train the AI,[27] it empirically countervails the argument that humans are no longer indispensable in the creative process. Hence, the question of whether AI-generated works are subject to copyright protection boils down to what extent of the work was a result of human intellectual effort.[28] While this question must be delved into on a case-by-case basis, technological advancements bestowed AI with more autonomy to makes its own decisions,[29] further blurring the distinction between an AI-generated work that bears human intellectual effort and one that is fully mechanical and autonomous without an iota of intellectual creativity exercised by human-being.
b) The Mystifying Question of Authorship
As a matter of law, the author of computer-generated LDMA works is ‘the person who undertakes the necessary arrangements for the creation of the work’.[30] As discussed above, authorship can only be found if the AI-generated work reflects substantial intellectual effort exercised by a human. The question of who is ultimately the author of the AI-generated work turns to what constitutes “necessary arrangement for the creation of the work”. Two categories of people appear to be the probable candidate. It can be the programmer of the AI system who facilitated the creation of the work in the first place by building and training the AI system. Equally, it can also be the user who instructed the AI system to create the work by entering keywords into search engine. However, while theoretically the question of whether the groups of programmers and users can co-exist or are mutually exclusive is yet to be clarified by Parliament, it is unlikely that the latter group can qualify as author given the practical ramifications of copyright protection being too non-exhaustive. As copyright ownership allows the owner to exclude other people from using it - imagine how many copyright claims would arise if people can sue one another just by the mere combinations of search phrases? Nonetheless, exceptions may apply if the instructions that one gives to the AI system sufficiently reflect the author’s free and creative “personal touch”[31] expressed through “choice, sequence and combination of words”.[32] However, this shadow of doubt is yet to be cleared by Parliament.
Normative Justifications and Conclusion
While it is being argued in this article that copyright exists in AI-generated works which reflect the author’s intellectual creativity in “making necessary arrangements to the creative process”, the full picture of copyright protection in AI-generated works will not be well-encapsulated without discussing the normative justifications of such copyright protection or the overarching question of whether AI-generated works should deserve copyright at all. From a utilitarian perspective, while copyright work created by a person usually takes months or years of effort and investment, it merely takes a few seconds for AI systems like ChatGPT to generate results and responses. The amount of time and effort invested in copyright works and computer-generated works, in essence, is disproportionate, granting the same level of incentive to these respective works will undoubtedly stifle and disincentivise human investment in a creation.[33] However, from a natural rights point of view, while copyright ought to protect the personality of authors expressed in their works, it appears incompatible with the protection of an AI-generated work which modified the original copyright work, as in the case of the AI-generated Drake and The Weeknd song. This risks stultifying creativity and disincentivising investment in creative projects, which ultimately goes against the underlying rationale of copyright protection. In addition, regarding wider policy considerations, the extent to which copyright should be given to computer-generated works is ultimately a balancing task of the competing interests between authors and users, so copyright protection should get accustomed to the climate of society in which AI have pervasively intruded into people’s daily life.
Harvey Mason Jr, the CEO of Recording Academy once said: “There is so much potential with AI but it also presents risks to our creative community so it is crucial that we [discuss this issue] right early on so we don't risk losing the artistic magic that only humans can create.”[34] To sum up, the controversy of whether copyright protection should subsist in AI-generated works continues to be a heated topic for public discourse be it within or beyond the creative industries and IP academics.
Photo credit: Unsplash
[1] Winnie Weng and Joey Lau, ‘Hong Kong: AI-generated content – who owns the copyright?’ Withersworldwide (10 July 2023) <https://www.withersworldwide.com/en-gb/insight/read/hong-kong-ai-generated-content-%E2%80%93-who-owns-the-copyright>
[2] s3(1); s4 CDPA
[3] Rita Matulioyte and Jyh-An Lee, ‘Copyright in AI-generated works: Lessons from recent developments in patent law’ (2022) 19 A Journal of Law, Technology & Society 1 <https://script-ed.org/article/copyright-in-ai-generated-works-lessons-from-recent-developments-in-patent-law/>
[4] Ibid.
[5] Andres Guadamuz, ‘Artificial Intelligence and Copyright’ WIPO Magazine (October 2017) <https://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html>
[6] Chloe Olewitz, ‘A Japanese A.I. program just wrote a short novel, and it almost won a literary prize’ DigitalTrends (23 March 2016) <https://www.digitaltrends.com/cool-tech/japanese-ai-writes-novel-passes-first-round-nationanl-literary-prize/>
[7] Dawn Chmielewski, ‘Black Mirror: Actors and Hollywood battle over AI digital troubles’ (14 July 2023) Reuters <https://www.reuters.com/business/media-telecom/union-fears-hollywood-actors-digital-doubles-could-live-for-one-days-pay-2023-07-13/>
[8] Andres Guadamuz, ‘Artificial Intelligence and Copyright’ WIPO Magazine (October 2017) <https://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html>
[9] Mark Savage, ‘AI-generated Drake and The Weeknd song goes viral’ British Broadcasting Corporation (17 April 2023) <https://www.bbc.co.uk/news/entertainment-arts-65298834>
[10] Ibid.
[11] Ibid.
[12] s3(1); s4 CDPA
[13] s5A-B CDPA
[14] s6 CDPA
[15] s8 CDPA
[16] s1(a) CDPA
[17] Infopaq Int v. Danske Dagblades Forening, Case C-5/08 [2009] ECR I–6569
[18] s178 CDPA
[19] s9(1) CDPA
[20] s9(3) CDPA
[21] Nova Productions v. Mazooma Games [2006] EWHC 24 (Ch); [2006] RPC (14) 379
[22] s9(3) CDPA
[23] Ibid.
[24] Luke Zaphir and Jason M Lodge, ‘Is critical thinking the answer to AI’, Times Higher Education (27 June 2023) <https://www.timeshighereducation.com/campus/critical-thinking-answer-generative-ai>
[25] [1964] 1 WLR 273
[26] Designers Guild v. Russell Williams [2000] 1 WLR 2416 (Lord Millett)
[27] Aaron Hayward, Anna Vandervliet, Byron Turner, Michael Dardis, Rachel Montagnon, Heather Newton Peng Lei, Alex Wang and Giulia Maienza, ‘The IP in AI: Does copyright protect AI-generated works?’ Herbert Smith Freehills (16 May 2023) <https://www.herbertsmithfreehills.com/insights/2023-05/the-ip-in-ai-does-copyright-protect-ai-generated-works>
[28] Ibid.
[29] Clifford Chance, ‘AI and IP: copyright in AI-generated works (UK law) Can copyright subsist in an AI-generated work?’ Clifford Chance (20 November 2017) <https://www.cliffordchance.com/insights/resources/blogs/talking-tech/en/articles/2017/11/ai-and-ip-copyright-in-ai-generated-works-uk-law.html>
[30] s9(3) CDPA
[31] EM Painer v Standard VerlagsGmbH, Case C‑145/10, EU:C:2011:798
[32] Infopaq Int v. Danske Dagblades Forening, Case C-5/08 [2009] ECR I–6569
[33] Winnie Weng and Joey Lau, ‘Hong Kong: AI-generated content – who owns the copyright?’ Withersworldwide (10 July 2023) <https://www.withersworldwide.com/en-gb/insight/read/hong-kong-ai-generated-content-%E2%80%93-who-owns-the-copyright>
[34] Mark Savage, ‘AI-generated Drake and The Weeknd song goes viral’ British Broadcasting Corporation (17 April 2023) <https://www.bbc.co.uk/news/entertainment-arts-65298834>