Picture a team of pioneering researchers in Nigeria embarking on an ambitious project. Using Machine Learning (ML) techniques, they dive into a vast collection of digital texts that encompass academic articles, literary works, and online content. Their mission? To unearth hidden patterns, glean insights, and unravel correlations that hold the potential to drive breakthroughs in various fields, ranging from medicine to the social sciences. While we’re at it, also imagine cutting-edge AI-powered platforms and applications, designed to analyze and generate innovative music inspired by the vibrant melodies of Afrobeats. a genre deeply rooted in Nigerian culture. These remarkable systems rely on an extensive dataset, including copyrighted Afrobeats tracks.[1]
Both scenarios illustrate endeavors fueled by the technique known as Text and Data Mining (TDM), which automates the processing of vast quantities of data. Research conducted by McKinsey & Company in 2011 revealed the remarkable potential of TDM, estimating its ability to generate significant value across sectors, such as $100 billion for healthcare service providers in the United States and up to $700 billion for global end users of personal location data.[2]As these involve the use of copyright works, the question arises: Should the researchers secure permission from the rightholders to employ these complete works as input for their AI and ML systems, rather than merely utilizing excerpts? Or should concerns solely arise if issues emerge in the output, such as significant similarities to pre-existing works employed for training?[3]
Against the foregoing background, this article explores the legality of TDM under Nigeria’s copyright law, discussing the perspectives on obtaining permission from rightsholders for using copyrighted material as input in AI/ML systems and examining the limitations of existing provisions in Nigeria’s Copyright Act. It also compares the approach to TDM in other jurisdictions and emphasizes the importance of explicitly providing a TDM exception in the copyright framework to promote research, innovation, and knowledge discovery while balancing the rights of creators and the needs of society.[4]
Arguments and Perspectives
On one hand, it can be argued that obtaining permission from rightsholders for using copyrighted material as input in AI systems is necessary to uphold the rights of creators and protect their intellectual property.[5] This perspective emphasizes the importance of respecting the exclusive rights granted to rightsholders under copyright law. By seeking permission, AI developers and users acknowledge and honor the rights of the original creators, ensuring that they are appropriately recognized and compensated for their work. This approach promotes a fair and equitable system that upholds the principles of copyright law while still allowing for technological advancements and innovation in the field of TDM.
On the other hand, it has also been argued that TDM should be seen as a technology that serves as a substitute for human viewing or reading, where the copying involved is merely incidental to the functioning of the technology.[6] After all, and as affirmed by the UK Supreme Court, viewing or reading a work does not constitute copyright infringement (Public Relations Consultants Association Ltd v The Newspaper Licensing Agency Ltd, [2013] UKSC 18). This perspective asserts that the information extracted through scanning through a copyrighted work should not be considered an infringement, regardless of whether it is done by humans or machines. However, infringement could potentially occur if there is substantial similarity between TDM outputs and pre-existing works in the training data.[7]
TDM Exception in Nigeria (Or the Lack Thereof?)
There is no explicit provision in Nigeria’s copyright law[8] addressing TDM. Section 20, which contains the general exceptions to copyright, includes provisions such as those on non-commercial research and private use (Section 20(1)(c)), and use for judicial or legislative proceedings (Section 20(1)(l)). These only touch on limited TDM (if at all).
Subsection (1)(o) of section 20 seems the most related to TDM. It provides that the rights conferred under sections 9 – 13 do not include the right to control any of the acts specified in those sections by way of fair dealing for purposes such as:
“transient and incidental reproductions, which are integral and essential parts of a technological process, whose sole purpose is to enable transmission in a network between third parties by an intermediary or for other lawful use, where such use has no independent economic significance”
This provision, which happens to be similar to the EU’s Article 5(1) of Directive 2001/29/EC (EU Copyright Directive), may be relevant to this issue. However, as with the EU law, it, too, is limited to TDM techniques that exclusively involve creating temporary reproductions that are “transient or incidental” and play a vital role in a technological process enabling lawful use without independent economic significance. Therefore, it is doubtful whether reproductions made for TDM purposes meet these requirements, particularly regarding the aspect of “transient” reproductions.
It has been argued, however, that Section 20 is an open general exception to copyright infringement, due to the inclusion of the phrase “such as” before the enumerated list of permitted purposes.[9] Consequently, it has been suggested that it could function similarly to the fair use right in the United States, allowing for the application of other purposes not explicitly stated in the provision.[10] This argument runs contrary to the rules of interpretation of statutes. The mere inclusion of the phrase “such as” before the enumerated list of permitted purposes neither creates a borderless scope of exceptions, nor makes Nigeria’s fair dealing provision equivalent to the fair use right in the United States. The presence of “such as” indicates that the listed purposes are examples, but the list provides a roadmap for what would fall within the scope and does not leave room for the inclusion of other purposes that do not easily fit within the category of the examples listed. Fair use on the other hand, provides a case-by-case analysis and allows for a wide range of transformative uses.
Also, it is admitted that the proviso in Section 20, which outlines the factors to be considered in determining fair dealing resembles the four-factor test used in the United States to assess fair use. Considering this, one could argue that Section 20 allows for a fair dealing provision that resembles the fair use concept, adding some level of flexibility in assessing whether a particular use qualifies as fair dealing. However, despite the similarities, it needs to be pointed out that fair dealing and fair use are distinct concepts with their own legal frameworks and interpretations. Their scope and application are more than likely to be interpreted differently by Nigerian and U.S. courts.
The interpretation and application of fair use in the United States have been shaped by decades of case law and legal precedents, providing a clearer understanding of its boundaries and scope. In Nigeria, the application and scope of Section 20 and the concept of fair dealing are yet to be extensively tested and defined through legal precedents. This lack of legal precedents and clear guidelines can lead to uncertainty and inconsistent outcomes when determining the applicability of other purposes not explicitly mentioned in Section 20.
While open general exceptions have been lauded for being able to accommodate unforeseen uses that are fair to the right holder, leaving TDM to be addressed solely under the fair dealing exception, without a specific provision in copyright law, can pose certain challenges. This can be burdensome and deter some legitimate TDM activities. Also, it has created uncertainty for researchers and organizations engaging in TDM, inhibiting innovation and cross-border collaborations. We are left speculating how the court would construe the Act’s fair dealing provisions in a scenario where AI-powered platforms and applications are developed to analyze Nigerian music and generate new music or Nollywood films based on copyright-protected works.
Flowing from the foregoing analysis, when copyrighted work, e.g., music, is intended to be used as input for AI systems generating new compositions, it is advisable to seek permission from rightsholders. This is unless the permission is already granted by default, as is the case with Creative Commons licenses, or if the work is in the public domain or considered orphaned.
The Status of TDM in Other Jurisdictions
Several jurisdictions, including Singapore, Japan, the European Union (EU),[11] and the UK, have implemented specific exceptions for TDM.[12] Under current UK copyright law, the text and data mining of copyrighted works for non-commercial purposes is permitted, provided that the user has lawful access to the work via, for example, a licence, subscription or permission in terms and conditions. Under U.S. copyright law, TDM is lawful due to the fair use provision. Carroll stated that researchers may lawfully conduct computational research on any scientific articles or data to which they have access so long as the durable outputs of this research do not incorporate more original expression than is permissible.[13]
By explicitly providing a TDM exception, policymakers can ensure legal certainty and clear boundaries for users, promoting TDM activities without the fear of infringement. Specific TDM exceptions recognize the benefits of TDM in advancing research and knowledge discovery while balancing the rights of creators and the needs of society. Empirical evidence shows that TDM research is better promoted through exceptions that more specifically authorize TDM research.[14]
Conclusion
Copyright law aims to strike a balance between the rights of creators and the needs of society. By explicitly providing a TDM exception, policymakers and lawmakers in Nigeria could have better ensured that the benefits of TDM, e.g., advancing research and knowledge discovery, are properly recognized and promoted. As Nigerian creative industries navigate the evolving landscape of TDM, it is crucial to recognize the implications and address the need for safeguards within the country’s copyright framework.
[1] Speaking of AI and music, this is already happening. See riffusion.com
[2] See McKinsey & Company. (2011). Big data: the next frontier for innovation, competition and productivity.
[3] Some have answered this question in the affirmative, arguing that such unauthorized and unremunerated use of their works negatively impacts the livelihood of authors. See for example: https://www.egair.eu/#manifesto accessed 14 June 2023; See also: The class action launched against Copilot, an AI trained to generate computer code using examples from Github.
[4] The impact TDM has on rightsholders is highlighted in a recent letter titled “Our Call for Safeguards Around Generative AI,” signed by authors, performers, and copyright holders across various creative industries in Europe. The said letter emphasized the pressing need to address concerns related to the usage of copyrighted material for training AI systems, along with the ensuing issues of accountability, liability, and fair remuneration.
[5] Eleonora Rosati (2019) Copyright as an obstacle or an enabler? A European perspective on text and data mining and its role in the development of AI creativity, Asia Pacific Law Review, 27:2, 198-217, DOI: 10.1080/10192557.2019.1705525
[6] Michael W. Carroll, ‘Copyright and the Progress of Science: Why Text and Data Mining Is Lawful’ (2019) 53 UC Davis L Rev 893; Maurizio Borghi, “Text & Data Mining”, Copyrightuser https://www.copyrightuser.org/understand/exceptions/text-data-mining/
[7] Read more on this here: João Pedro Quintais, “ Generative AI, Copyright and the AI Act”, Kluwer Copyright Blog, 9 May 2023: https://copyrightblog.kluweriplaw.com/2023/05/09/generative-ai-copyright-and-the-ai-act/ accessed 14 June 2023
[8] Nigerian Copyright Act, 2022
[9] Dick Kawooya and Sean Flynn, “Comments to Nigerian copyright reform consultation”, 10 October 2021: https://digitalcommons.wcl.american.edu/cgi/viewcontent.cgi?article=1030&context=pijip-righttoresearch-testimony
[10] Flynn, Sean and Palmedo, Michael and Izquierdo, Andres, Research Exceptions in Comparative Copyright Law (2021). PIJIP/TLS Research Paper Series no. 72, Available at SSRN: https://ssrn.com/abstract=3961017
[11] This article discusses a few of EU’s TDM exceptions: https://copyrightblog.kluweriplaw.com/2023/05/09/generative-ai-copyright-and-the-ai-act/
[12] Ibid
[13] Michael W. Carroll, ‘Copyright and the Progress of Science: Why Text and Data Mining Is Lawful’ (2019) 53 UC Davis L Rev 893
[14] See Christian Handke et al., Is Europe Falling Behind in Data Mining? Copyright’s Impact on Data Mining in Academic Research, in New Avenues for Electronic Publishing in the Age of Infinite Collections and Citizen Science: FfScale, Openness and Trust 120–130 (Brigit Schmidt & Milena Dobreva eds., 2015) cited in Dick Kawooya and Sean Flynn, Comments to Nigerian copyright reform consultation, 10 October 2021: https://digitalcommons.wcl.american.edu/cgi/viewcontent.cgi?article=1030&context=pijip-righttoresearch-testimony
Leave a Reply