Copyright in the age of artificial intelligence: Navigating access to algorithmic training materials and the three-step test for text and data mining in Nigeria

Abstract

Over the past decade, the Nigerian government has sought to leverage Artificial Intelligence (AI) to drive socio-economic transformation and improve the welfare of its citizenry. Recent initiatives, such as the establishment of the National Centre for AI and Robotics (NCAIR) and the development of several strategic AI policies, highlight the country's commitment to this objective. This article explores the often-overlooked issue of how the Nigeria's copyright regime hinders these initiatives, revealing that the regime permits only fair dealing and the transient or incidental reproductions of copyrighted materials for limited technological purposes. This study argues that this regime is unduly restrictive for algorithmic training and risks stifling AI innovation and the development of machine-learning models in Nigeria. It recommends adopting a bespoke text and data mining (TDM) exception tailored to Nigeria's needs, allowing the use of copyrighted works for training AI models and machine learning activities within defined limits. Drawing on comparative analyses of copyright frameworks in jurisdictions such as Singapore, Japan, the United Kingdom, and the European Union, this study demonstrates that the proposed TDM exception aligns with the three-step test under international copyright conventions. For instance, the exception is limited to specific users and types of reproductions, applies only to internalized and transformative reproductions, and avoids traditional methods of exploiting copyrighted works that prejudice the legitimate interests of rightsholders. The ultimate goal of this exception is to recalibrate Nigeria's copyright system to justly balance AI innovation with authors' rights, aligning it with foundational principles of the international copyright system in an era of rapid technological advancements.