Author: Manish Kumar
A DPIIT-appointed committee has submitted a Working Paper on Generative AI and Copyright – Part I, in December 2025, which investigates the legal and policy implications of using copyrighted works as training data for generative artificial intelligence (GenAI). It examines India’s existing Copyright Act, 1957 highlighting the absence of a text-and-data mining (TDM) exception under Indian law. Further, the Paper explores global regulatory approaches and synthesises stakeholder consultations to assess competing models for access to training data and remuneration of right-holders. The Paper is limited to input-side concerns relating to the use of copyrighted content in AI training, while Part II will address authorship and protection of AI-generated outputs.
The Paper identifies a fundamental conflict between the scale and diversity of data required to develop GenAI systems and the risk that unlicensed ingestion of copyrighted works could undermine the sustainability of creative industries. The paper raises two questions central to the issue; 1) whether training AI Systems involves exercising any of the exclusive rights given to copyright holders under Section 14 of the Copyright Act, and 2) whether the “fair dealing” exception under Section 52(1)(a) of the Copyright Act, 1957, can be interpreted to protect the training of AI Systems.
The Paper further pinpoints the sharply divergent views of various Stakeholders: 1) the AI developers and technology firms that advocate exceptions or simplified access regimes, 2) the authors, publishers, and other right-holders emphasise licensing mechanisms to protect control and remuneration. The Paper undertakes a comparative review of international approaches, including blanket TDM exceptions, opt-out regimes, voluntary and collective licensing systems, and statutory compensation models, as well as ongoing litigation in various jurisdictions. It concludes that none of these approaches, when adopted in isolation, adequately address India’s legal structure and creative ecosystem.
The Paper explains that acts intrinsic to AI training, such as reproduction, storage, adaptation, and communication of works, are presently treated as exclusive rights under the Copyright Act, 1957, with no exception permitting TDM. The committee proposes a hybrid statutory framework described as “One Nation, One License, One Payment”, under which AI developers would be granted a mandatory blanket licence to use lawfully accessed copyrighted works for training purposes without requiring individual authorisation. Creators would receive statutory remuneration, with royalties collected and distributed by a single government-designated, non-profit body comprising existing copyright societies and collective management organisations – CRCAT (Copyright Remuneration Collective for AI Training). The framework envisages transparent royalty pooling, distribution to members and non-members of CRCAT on registration for the purpose of receiving royalties for AI training, institutional mechanisms for rate-setting, judicial review, and dispute resolution. The Paper also acknowledges implementation challenges in the distribution of royalties, highlighting the need to ensure transparency and accountability.
Disclaimer: Views, opinions, interpretations are solely those of the author, not of the firm (ALG India Law Offices LLP) nor reflective thereof. Author submissions are not checked for plagiarism or any other aspect before being posted.
Copyright: ALG India Law Offices LLP