Copyright In The Age Of Generative AI: Legal Challenges And The Fair Use Doctrine In The Context Of Training Data
- IJLLR Journal
- Oct 14
- 1 min read
Prof. (Dr.) Asmita Vaidya, Principal, Government Law College, Mumbai
M.E.S.V. Krupakar, Research Scholar, Dept. of Law, University of Mumbai
ABSTRACT
The rapid proliferation of generative artificial intelligence (AI) systems has intensified legal and ethical debates concerning the use of copyrighted material in training datasets. These systems are built on large-scale ingestion of digital content like text, images, and audiovisual works which are sourced from publicly accessible platforms without express authorization. Such practices raise complex questions regarding unauthorized reproduction, derivative works, and the limits of permissible use under copyright law.
This paper examines the legal ramifications of incorporating copyrighted content into AI training, with particular emphasis on the fair use doctrine in the United States, while also considering parallel fair dealing frameworks in other jurisdictions. It traces the historical development of fair use and its expansion to accommodate technological innovations, especially transformative and non-expressive uses, through landmark decisions.
The paper further engages with contemporary scholarly debates and doctrinal tensions surrounding the application of fair use to AI training. It provides a detailed analysis of recent litigation involving Bartz v. Anthropic, Kadrey v. Meta, andThomson Reuters v. ROSS highlighting jurisprudential inconsistencies both across these cases and in relation to established precedent on transformative use. Ultimately, the paper argues that in light of disruptive technological advances such as AI, legal systems require greater predictability and clarity and the adoption of explicit text and data mining exceptions may offer a more coherent and future-ready framework for balancing innovation with the protection of authors’ rights.
Keywords: Generative AI, Copyright Law, Fair Use Doctrine, Training Data, Text and Data Mining (TDM)
