Authorship, Ownership and Liability in AI-Generated Works: A Critical Analysis of Human Authorship and Training Data Regimes in Generative AI

Abstract

The advanced growth of generative artificial intelligence has fundamentally challenged traditional dogmas of intellectual property (IP) law, especially the concepts of authorship, ownership and liability. The legal issues raised by AI-generated works that are critically analyzed in this paper are the changing definition of human authorship and the controversial concept of training data regimes. The period of 2023 to 2025 such as the case of Thaler v. Perlmutter (2023-2025) and the U.S. Copyright Office AI guidelines, and the litigations of Getty Images v. Stability AI and The New York Times v. OpenAI indicate that the judiciary continues to insist on human creativity as a precondition to receiving copyright protection while disclosing the unresolved tensions in the use of copyrighted data in training AI.

The research claims that the current IP structures, based on anthropocentric conceptualization of creativity, become less effective to deal with practices of algorithmic production of content and massive ingestion of data. It also questions the fair use, consent, and liability doctrines available today whether they are doing enough to balance the innovations with the rights of producers, particularly in an area such as India where the regulation remains unclear. The paper focuses on the emergent global response to the issue by taking a comparative and doctrinal look at the EU AI Act and policy discussions regarding data governance.

The paper has concluded that the IP law needs a recalibration, one that redefines the authorship criteria, defines the ownership rights, sets clear and fair standards of training data to make sure not only the technological progress but also legal responsibility is present in the era of generative AI.