Read, Learn, Infringe? The Legal Paradox Of AI Training Data Under Indian Copyright Law
- IJLLR Journal
- 1 hour ago
- 1 min read
Rutuja Bhand, LL.M. (IP and Tech Laws), Vishwakarma University, Pune
Rahi Ajabe-Alhat, Assistant Professor of Law, Vishwakarma University, Pune
ABSTRACT
Big AI models need tons of information to grasp how words, pictures, and trends work. From online sources, huge chunks get pulled without human help - books under copyright, news pieces, photos, lines of code, movies, artistic stuff. Here’s the twist: these machines depend on duplication just to study, but making copies like that might break ownership rules. Places like the U.S., EU, and UK are shaping laws to handle this clash. Not so in India - no firm answer yet on if using protected content to train AIs counts as allowed.
Back when India made its Copyright Act in 1957, artificial intelligence wasn’t even a thought. Because of that timeline gap, rules about copying, violations, or permitted use lack clarity on modern actions like pulling data online, building training sets, or teaching machines.
This study looks at parts 14, 51, 52, and 2(d) of India’s Copyright Act, alongside court rulings including Eastern Book Company v DB Modak and RG Anand v Deluxe Films. Meanwhile, practices in the United States, Europe, and Britain are reviewed, especially how they handle limited usage, scanning material for patterns, and openness rules.
A fresh look at the findings shows India needs fairer rules for handling AI training data. These guidelines must support progress without ignoring those who create content.
Keywords: AI; Copyright; Fair Dealing; Data Mining; Copyright Violation; Indian Copyright Act; AI Regulation; Machine Learning; Generative; Intellectual Property Law.
