The Input Side Crisis AI Training Data and the Future of Fair Use in India Hemanth Kumar

The "Input Side" Crisis: AI Training Data and the Future of Fair Use in India

Submitted by

Uday Kiran (Student)

Sanjay N (Student)

Under the Guidance of

Dr. Gitanjali (Faculty, IFIM College)

Department of Commerce and Management IFIM College, Bengaluru, India

Submitted To International journal of contemporary legal research

The "Input Side" Crisis: AI Training Data and the Future of Fair Use in India

Abstract

The ingestion of copyrighted works for Artificial Intelligence (AI) training has precipitated an "input side" crisis in India, challenging the structural integrity of the Copyright Act, 1957. This paper argues that the current fair dealing framework under Section 52(1)(a) is doctrinally ill-equipped to subsidize commercial AI ingestion. While developers characterize training as "non-expressive" use, Indian jurisprudence—anchored by Eastern Book Company v. D.B. Modak and intensified by recent litigation in ANI Media v. OpenAI—demands a nexus of human creativity that automated scraping fails to satisfy.Central to this study is an analysis of the DPIIT’s December 2025 Working Paper, which proposes a "Hybrid Model" of statutory licensing. This "One Nation, One License, One Payment" framework shifts the paradigm from unremunerated "opt-out" systems toward a mandatory Copyright Royalties Collective. The paper further evaluates the MeitY 2025 AI Governance Guidelines, specifically the "Graded Liability" and "Traceability" mandates for training datasets. By synthesizing these regulatory shifts, the research demonstrates how India is pioneering a "Rights-Centered" architecture that rejects Western "Fair Use" flexibility in favor of protecting cultural heritage and creative labor. Ultimately, the study proposes a legislative roadmap to balance high-quality data access with the constitutional mandate to safeguard the economic rights of creators in a borderless digital ecosystem.

Keywords: Artificial Intelligence; Section 52(1)(a); Hybrid Model Licensing; DPIIT Working Paper (2025); ANI Media v. OpenAI.