AI AND COPYRIGHT: FROM A DOCTRINAL CRISIS TO A HYBRID MODEL OF COLLECTIVE LICENSING

Abstract

The explosive development of generative artificial intelligence (hereinafter –  AI) has triggered a fundamental crisis within intellectual property law. A central conflict has emerged between artificial intelligence developers, who require access to vast datasets for model training, and rights holders, who seek to protect their works and receive fair compensation for their use. The purpose of this article is to demonstrate the inadequacy of existing legal doctrines in addressing the use of copyrighted data for training generative AI and to substantiate the need for a transition to a hybrid model of compulsory collective licensing as the most rational alternative. The research is based on a comprehensive application of general scientific and special legal methods. The formal-legal method was used to analyze the norms of United States law and European Union (hereinafter – EU) law. The comparative-legal method allowed for the identification of common features and key differences between the American and European approaches and an assessment of their relative effectiveness. Systemic analysis was applied to view copyright as an integral system and to identify the fundamental challenges posed by artificial intelligence to its core elements (authorship, originality, economic incentives). The synthesis method enabled the generalization of the findings and the formulation of the author's hypothesis. The article proves that neither the flexible but unpredictable fair use doctrine nor the formalized but ineffective Text and Data Mining (hereinafter – TDM) exception can resolve the issue. As a solution, the architecture of a hybrid model of compulsory collective licensing is proposed. This research affirms that a continued reliance on legacy legal doctrines creates a landscape of perpetual instability. The case-by-case litigation engendered by the fair use doctrine proves to be an inefficient and unpredictable means of governance, while the EU's TDM opt-out mechanism has been shown to be functionally incapable of providing meaningful compensation to creators.