Zeynep Ayça ŞENOL
Bilkent Üniversitesi Hukuk Fakültesi
Generative Artificial Intelligence relies on machine learning, which is an AI type (Martineau). Through machine learning, generative AI models “absorb” massive amounts of raw data to be able to mimic human cognition. Owing to the training, AI models develop in a way that they do not need a step-by-step walkthrough of instructions to be able to generate the prompt that they receive; they learn to operate with no human interruption in the end (Guadamuz). Developers embed algorithms that enable the constant training, evolution, and decision-making processes into AI models. Recently, there has been an observably increasing trend among developers to tailor machine learning algorithms to generate visual arts materials like paintings and photographs, music, and literary pieces. AI tools like Dall-E, ChatGPT, and Midjourney seem to lead the way in this regard.
“Global” copyright protection has been ensured mainly through three multilateral texts. Human expression in literary and artistic works has been protected through copyright regulations under the Berne Convention, dated 1886. The Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS Agreement) of 1995 foresees copyright protection relating to new technologies like computer programs and databases. The WIPO Copyright Treaty (WCT), dated 1996 ensures the distribution, making publicly available, and rental rights of creative works (WIPO). Aside from the mentioned texts, individual countries have implemented certain IP laws to regulate the matter. While the Common Law tradition acknowledges the display of a minimum amount of authorial judgment, labor, and skill, the Civil Law tradition demands a stronger bond; the creative work needs to bear a stamp of the creator’s personality and creative efforts, which clearly requires more than mere judgment, labor, and skill (WIPO). Nevertheless, the two strongly agree on one element, which is to motivate and enthrall human creativity through legal protection. In essence, the eligible end products of human creativity and cognition receive consideration under current copyright protection systems. Up until 2019, IP offices have received more than 340,000 AI-related applications according to World Intellectual Property Organization Technology Trends on Artificial Intelligence, that is much earlier than when generative AI tools got mainstream among regular masses (WIPO Secretariat).
The legal status of granting copyright protection to AI-generated works is not clearly codified. Nevertheless, IP laws of many countries prioritize the involvement of human cognition in a creative manner (Guadamuz). Members of The International Association for the Protection of Intellectual Property (AIPPI) have split views on whether AI-generated works should be granted copyright protection or not. AIPPI’s resolution signifies the need for human intervention and intelligence in the active creation of the generated work along with an originality element (Intellectual Property Office UK).
In the monumental Infopaq International v. Danske Dagbaldes case, the Court of Justice of the European Union (CJEU) relied on its 2001 Directive on the Harmonisation of Certain Copyright and Related Rights in the Information Society and ruled that during data capturing, printing out words as the outcome of searching through newspaper articles does not comply with the said Directive; the Directive ensures protection to works with “the expression of the intellectual creation of the author of the work”, thus the process depends on the articles’ rightholders consent (CJEU). Similarly, the US Supreme Court’s landmark decision on Feist Publications v. Rural Telephone Services highlights the originality element for copyright protection according to the Constitution and the Copyright Act and ensures that any information not involving an originality and creativity element is not copyrightable (US Supreme Court). In 2023, the US Copyright Office required the reissuing of the copyright protection application of the graphic novel “Zarya of the Dawn”. The Office set forth its stance in its letter indicating that the written body and the arrangements of the images by the author of the novel are copyrightable, not the entire creation that includes illustrations created by the AI program Midjourney. Established case law of the US courts has been exhibiting the interpretation of the “works of authorship” as human authors’ creations (Taylor English). Evidently, artists’ and authors’ main concern regarding copyright infringement is the fact that the AI models use their work as training material to produce what the prompts command. Artists McKernan, Andersen, and Otiz filed a lawsuit against Stable Diffusion that used thousands of protected works of artists to generate collages (US District Court).
Through an analysis of the works of Rembrandt Harmenszoon van Rijn, a 17th-century Dutch painter, a computer generated the portrait “The Next Rembrandt” in 2016. The painting drew out sponsorships and collaborations of high economic value including ING Banking Group and Microsoft (Microsoft News). Such situations bring out the question of how fair it would be even for creations with such considerable commercial values to remain completely unprotected. The UK Copyrights, Designs, and Patents Act (CDPA) of 1988 Section 9 Article 3 vividly states, “In the case of a literary, dramatic, musical or 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.” enabling the programmers to be the copyright holders of such (UK Public General Acts).
The standalone Chinese case Li v. Liu has declared copyright protection to an AI generated image on three bases. The image complied with the Chinese Copyright Law by fulfilling originality and being an artwork criteria as it carried intellectual and artistic choices of the user. Secondly, the user got considered as the author and not the company since the user’s aesthetic perspective directed the model. Thirdly, because the defendant was found to be liable for infringement as they removed the copyright mark on the image (Guadamuz).
Before arguing whether creations of AI should be copyrighted, it should be recognized that copyright is a coin with two sides. Copyright as the umbrella covers and protects two types of rights, economic rights including right of reproduction, right of making available to the public, and rights of adaptation and translation; moral rights including attribution rights and integrity rights. For the commercial purposes of copyright protection it is important to recognize that economic rights are transferable whereas moral rights are non-transferable (WIPO). Through moral rights, creators get to assert their authorship along with their integrity and reputation against infringements and misuse. Therefore, against any derogatory output that an AI model creates as per the prompt it receives, the developer has the right to restore creative integrity. Nevertheless, situations as such arise a dilemma. In my opinion, even though developers as real persons or companies as legal persons own the AI models, their ownership shall not be extending over users’ desired outputs. Otherwise, it would be like Microsoft Word owning the intellectual rights of every document ever created. Nevertheless, since the developers or companies provide artificial cognition with creative capability for users rather than a mere medium for an output, co-ownership should be considered as a reasonable and fair solution. After all, what the AI models come up with requires the creative input of both the developers/owners and the users.
The European Parliament’s Report on Intellectual Property Rights for the Development of Artificial Intelligence Technologies sets out numerous possible amendments. The Report envisages the need to distinguish between IP rights for AI-generated content and the development of AI technologies: to discover the potential of AI technologies in education and culture “unnecessary legal barriers” need to be removed. Moreover, the Parliament stresses the importance of a shift in the understanding of the creativity element, proposing that the creative result shall be the basis for an AI-generated creation to be considered as a work of art in lieu of its creative process. The report adopts a different appeal regarding the innovative and commercial outcomes of AI creations: the discernment of the deviation between autonomously generated creations and human creations with AI assistance is held vital (Séjourné).
Through a series of redefinitions and revisions of “creativity” within legal frameworks, the UK’s CDPA seems to be a suitable and applicable signpost to follow. Provided that both get protection, perhaps through the remedy of shared ownership, differentiating between a completely human-made creation and a generative AI-based creation in the legislative process could help restore the much-needed equity within the field. After all, IP Law is a means of providing economic assurance to creators through a legal lens regardless of the region. It seems like people do not need to rage against the machine anymore, all they need is skillful IP lawyers by their side.
Court of Justice of the European Union. Infopaq International A/S v Danske Dagblades Forening. 16 Jul. 2009, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A62008CJ0005. Accessed 7 Dec. 2023.
Guadamuz, Andres. “Artificial Intelligence and Copyright.” WIPO Magazine, 2017, https://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html. Accessed 7 Dec. 2023.
Guadamuz, Andres. “Chinese Court Declares that AI-Generated Image has Copyright” TechnoLlama. 9 Dec. 2023, https://www.technollama.co.uk/chinese-court-declares-that-ai-generated-image-has-copyright. Accessed 10 Dec. 2023.
Intellectual Property Office UK. “Artificial Intelligence Call for Views: Copyright and Related Rights.” GOV.UK, the UK Government, 23 Mar. 2021, www.gov.uk/government/consultations/artificial-intelligence-and-intellectual-property-call-for-views/artificial-intelligence-call-for-views-copyright-and-related-rights.
Martineau, Kim. “What Is Generative AI?” IBM Research Blog, IBM, 18 Aug. 2023, research.ibm.com/blog/what-is-generative-AI.
Microsoft News Office Europe. “The Next Rembrandt.” Microsoft News, 13 Apr. 2016, https://news.microsoft.com/europe/features/next-rembrandt/. Accessed 8 Dec. 2023.
Séjourné, Stéphane. “Report on Intellectual Property Rights for the Development of Artificial Intelligence Technologies.” The European Parliament, 2 Oct. 2020, https://www.europarl.europa.eu/doceo/document/A-9-2020-0176_EN.html#_section7.
Taylor English Duma LLP. “Re: Zarya of the Dawn.” Received by Van Lindberg, United States Copyright Office, 2021, https://www.copyright.gov/docs/zarya-of-the-dawn.pdf. Accessed 7 Dec. 2023.
UK Public General Acts. “Copyright, Designs and Patents Act 1988.” legislation,gov.uk, the UK Government, 31 Nov. 2003, https://www.legislation.gov.uk/ukpga/1988/48/section/9.
US Supreme Court. Feist Publications v Rural Telephone Service Company, Inc. 27 Mar. 1991, https://www.law.cornell.edu/supremecourt/text/499/340. Accessed 2023.
US District Court for the Northern District of California. Andersen et al v. Stability AI Ltd. et al. 13 Jan. 2023, https://dockets.justia.com/docket/california/candce/3:2023cv00201/407208. Accessed 4 Dec. 2023.
WIPO Secreteriat. “WIPO CONVERSATION ON INTELLECTUAL PROPERTY (IP) AND ARTIFICIAL INTELLIGENCE (AI) SUMMARY OF SECOND AND THIRD SESSIONS.” Geneva, 8 Jan. 2021.
WIPO. “DL101 Copyright.” WIPO Distance Learning Center. Accessed 5 Dec. 2023.