The Input and Output Considerations of Generative Artificial Intelligence and Copyright Infringement

1.0 Introduction

Generative Artificial Intelligence (GenAI) represents a revolutionary power that generates text alongside images as well as music and videos by requiring minimum human involvement. AI systems have reached considerable development levels in autonomous copyrightable work generation which demonstrates output capabilities that go beyond traditional limitations. Artificial Intelligence systems that use generative models and neural networks now generate creative outputs without any human assistance. For example, AI-generated visual art has made headlines, such as the portrait “Edmond de Belamy”^[1] created by the Paris-based art collective Obvious using Generative Adversarial Networks (GANs). The sale of this artwork at auction demonstrated AI’s competency in generating marketable creations worth commercial value^[2]. OpenAI MuseNet alongside Google Magenta demonstrate AI capacity to generate unique music compositions throughout different musical styles thus displaying AI compositional capabilities for sophisticated innovative compositions^[3]. The Generative Pre-trained Transformer 3 technology (GPT-3) produces text outputs with coherent sequences and contextual relevance that resembles human-written works making human judgement difficult^[4].

AI’s expanding role in copyrightable work production proves transformative for creative industries while generating important questions about authorship and copyright protection. AI’s technological capabilities increase efficiency and creativity yet create fundamental legal concerns about copyright law framework. The infringement of copyright occurs when anyone uses or reproduces or distributes or presents a copyrighted piece without obtaining authorization from its legal owner^[5]. Users may violate copyrights through text duplication or music or graphic theft along with developing new works from existing content. Copyright law in Nigeria like other jurisdictions establishes protections for original works of authorship yet raises uncertainties about AI-generated content rights and AI utilisation of copyrighted materials.

Precious Azubuike (Associate, HARLEM)

This article explores the input and output considerations of GenAI in the context of copyright infringement, analyzing how AI systems interact with existing copyrighted works and the potential liabilities for creators, AI developers, and users. It also examines the Nigerian legal framework alongside international perspectives, providing insights into possible regulatory approaches.

2.0 The Input Considerations of Generative AI and Copyright

The input considerations of generative AI and copyright relate to copyright issues as to the kind of data used to train generative artificial intelligence by its developers. Three key issues arising from the input considerations of generative AI are raining data and copyrighted works, and fair use and fair dealing considerations.

Training Data and Copyrighted Works

Generative AI models require vast amounts of data to learn patterns and generate human-like content. These datasets often include copyrighted texts, images, and music. The key question is whether training an AI model on copyrighted material without permission constitutes copyright infringement. This can be exemplified by current legal controversies like Getty images v Stability AI^[6] where Getty Images filed a lawsuit against Stability AI, alleging that the company used Getty’s copyrighted images to train its AI model without permission. Getty argued that this use constituted copyright infringement as it involved unauthorized reproduction and use of their images.

Under Nigerian copyright law, the Copyright Act 2022 provides protection to literary, musical, artistic, and cinematographic works, granting exclusive rights to authors, including reproduction and adaptation rights.^[7] Similar to international frameworks, Nigerian law does not explicitly address AI-generated content. However, unauthorized use of copyrighted material for AI training could violate reproduction rights.^[8]

Fair Use and Fair Dealing Considerations

Amidst all these controversies, perhaps the most contentious issue regarding the fair dealing exception is the possibility of applying it in training of artificial intelligence. Some jurisdictions, such as the United States, apply the “fair use” doctrine, which allows limited use of copyrighted works for purposes like criticism, research, or education.^[9] In contrast, the United Kingdom and Nigeria follow the “fair dealing” doctrine, which is more restrictive and permits specific uses, such as private study, research, or news reporting^[10]. Whether AI training qualifies as fair dealing remains uncertain under Nigerian law.

The argument is that if an AI training is to be considered research then it comes under the fairness exception of fair dealing^[11]. This argument is based on the fact of whether the training of AIs is acceptable as an actual exercise of research or simply a business venture with the aim of making profit.

Besides, the quantity of use, the fairness of use is another essential factor, and using entire works or substantial parts of them in AI training may be considered unfair especially where generated output compete with the original works in the market^[12]. While AI companies may argue that training models on copyrighted works falls under fair use, rights holders contend that mass data scraping without permission is an infringement.^[13] This legal ambiguity necessitates clearer guidelines in Nigeria’s copyright framework.

3.0 The Output Considerations of Generative AI and Copyright

The output considerations of generative AI deals with the copyright issues arising out of the output generated by AI.

Authorship and Ownership of AI-Generated Content

A fundamental question in copyright law is whether AI-generated works qualify for protection and, if so, who holds the rights. Traditional copyright law requires human authorship^[14]. In the United States, the Copyright Office has rejected copyright claims for AI-generated works without human input.^[15] This is consistent with Naruto v Slater case^[16], in which a court in the United States found that a monkey can not even own a copyright in photos that it took, this affirmed the principle that only living beings can be considered authors. In this case, a lawsuit was filed by animal rights activists on behalf of a macaque monkey named Naruto, who had taken a selfie using a camera owned by wildlife photographer David Slater. The case involved the question of whether Naruto had copyright ownership of the photograph he took. The court ruled that animals cannot hold copyrights. This implies that the authorship rights cannot be afforded to a non-human entity such as AI.

Similarly, the UK Copyright, Designs and Patents Act 1988 attributes copyright in computer-generated works to the person who made the necessary arrangements^[17]. In Nigeria, copyright law does not explicitly address AI-generated works, leaving uncertainty about whether a developer, user, or AI itself could claim authorship. The absence of legal recognition for non-human creators means that AI-generated works might fall into the public domain unless a human makes substantial creative contributions.^[18]

Derivative Works and Substantial Similarity

AI-generated content may closely resemble copyrighted works, leading to potential infringement claims. Courts assess infringement based on “substantial similarity” to existing works.^[19] If an AI model produces an artwork that closely mimics a copyrighted piece, it could be deemed an infringing derivative work, violating the rights of the original author^[20].

In Nigeria, courts may apply the test of substantial similarity to determine whether AI-generated content constitutes infringement. However, given the novelty of AI-generated works, a case-by-case approach would be required.

Moral Rights and Attribution Issues

Moral rights protect an author’s right to be credited for their work and prevent distortion of their creations.^[21] AI-generated content complicates attribution, as it is unclear who deserves credit—the developer, the user prompting the AI, or the AI itself^[22]. Nigerian copyright law recognizes moral rights, meaning AI-generated modifications of human works could infringe upon an author’s right to integrity.

4.0 Legal and Policy Implications

Judicial and Legislative Responses

Since Nigerian copyright law does not yet explicitly address AI-generated content, courts may rely on existing principles while awaiting legislative reforms. Internationally, some countries are considering regulatory frameworks to address AI’s copyright implications. For example, China’s 2023 guidelines require AI-generated content to respect intellectual property rights^[23]. Nigeria could adopt similar policies to clarify liability for AI developers and users.

 Proposed Solutions

To balance AI innovation with copyright protection, Nigeria should consider the following:

• Clarifying Fair Dealing Exceptions: Explicitly define whether AI training falls within fair dealing provisions.
• Establishing Licensing Mechanisms: Require AI developers to obtain licenses or compensate copyright holders for training data.
• Recognizing AI-Generated Works: Define conditions under which AI-assisted creations can qualify for copyright protection.
• Developing AI Attribution Standards: Introduce rules for crediting AI-generated works to ensure compliance with moral rights.

5.0 Conclusion

Generative AI presents complex copyright challenges, particularly regarding its use of copyrighted input data and the status of its outputs. While Nigerian law does not yet provide clear guidelines, existing copyright principles suggest that unauthorized AI training and substantially similar AI-generated works could constitute infringement. Legal reforms are needed to address AI’s impact on copyright law, ensuring a fair balance between innovation and intellectual property rights. As AI continues to evolve, proactive regulatory measures will be essential in shaping Nigeria’s copyright landscape in the digital age.

FOOTNOTES:

^[1] CNN (October 25, 2018), A sign of things to come? AI-produced artwork sells for $433K, smashing expectations. Retrieved from https://www.google.com/amp/s/amp.cnn.com/cnn/style/article/obvious-ai-art-christies-auction-smart-creativity

^[2] Stephensen, Jan Løhmann. “Towards a Philosophy of Post-creative Practices?–Reading Obvious’“Portrait of Edmond de Belamy”.” Politics of the Machine Beirut 2019. BCS Learning & Development, 2019.

^[3] Chauhan, Aditya. “Redefining Artistic Boundaries: The Impact of Generative AI on Creative Processes and Innovation.” (2024).

^[4] Roumeliotis, Konstantinos I., and Nikolaos D. Tselikas. “Chatgpt and open-ai models: A preliminary review.” Future Internet15.6 (2023): 192.

^[5] Lionel Bently, Brad Sherman, Dev Gangjee, and Phillip Johnson, Intellectual Property Law (Oxford University Press, 2022) 123.

^[6] Getty Images (US), Inc. v Stability AI, Inc, No. 23-cv-00201 (SDNY 2023).

^[7] Copyright Act 2022 (Nigeria), s.1.

^[8] Ibid., s.6.

^[9] Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569 (1994).

^[10] Copyright Act 2022 (Nigeria), s.21.

^[11] John McCarthy, ‘What is Artificial Intelligence?’ (2007) Stanford University.

^[12] Simon McDonald, ‘AI and the Future of Creativity’ (2019) 9(7) Tech Innov Manag Rev 24, 25.

^[13] J. O. Asein, Nigerian Copyright Law & Practice (2nd edn, Books and Gavel 2020) 145

^[14] Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991).

^[15] U.S. Copyright Office, Compendium of U.S. Copyright Office Practices (3rd edn, 2021) §313.2.

^[16] Naruto v Slater [2016] 247 F Supp 3d 1112 (ND Cal).

^[17] Copyright, Designs and Patents Act 1988 (UK), s.9(3).

^[18] Iaia, Vincenzo. “To be, or not to be… Original under copyright law, that is (one of) the main questions concerning AI-produced works.” GRUR International 71.9 (2022): 793-812.

^[19] Harper & Row v. Nation Enterprises, 471 U.S. 539 (1985).

^[20] P. Samuelson, “Generative AI and Copyright Law” (2023) 68 Journal of Law and Innovation 112.

^[21] Copyright Act 2022 (Nigeria), s.17

^[22] M. Rimmer, “Who Owns AI-Generated Content?” (2023) 45(1) Modern Law Review 112.

^[23] China National Copyright Administration, Guidelines on Copyright Protection for AI-Generated Content (2023).