The rapid advancement of artificial intelligence (AI), particularly large language models and generative AI systems, has transformed the production, dissemination, and commercial use of knowledge in the digital era. These systems rely on vast datasets drawn from books, research articles, websites, and other digital materials, many of which may be protected by copyright. While such datasets have enabled technological innovation, they have also raised serious legal and ethical concerns regarding authorization, compensation, ownership of training data, and the commercialization of intellectual works. This paper examines the phenomenon of “copyright capture,” whereby large technology corporations derive economic value from copyrighted materials through AI training without clearly defined licensing or benefit-sharing mechanisms. The issue is intensified by the dominance of companies possessing the infrastructure, computational capacity, and financial resources required to develop large-scale AI models. At the same time, academic publishers continue to restrict access to scholarly literature through subscription-based paywalls, creating a paradox in which AI systems may process knowledge that remains inaccessible to many human researchers, particularly in developing countries and under-resourced academic institutions. Using a doctrinal legal research methodology, the paper analyses copyright principles, academic publishing practices, the TRIPS Agreement, and emerging AI governance developments. It argues that existing international copyright regimes inadequately address AI training datasets, automated data extraction, and AI-generated outputs. The paper concludes that transparent data governance, fair licensing frameworks, stronger open access policies, and coordinated international cooperation are necessary to balance technological innovation with creators’ rights and equitable access to knowledge in the AI age.