Reconceptualising the Concept of Copyright Law with Special Regard to Training of Artificial Intelligence in India

Abstract

Artificial Intelligence systems are trained by ingestion of vast datasets of information and computational analysis of copyrighted work without direct involvement of human beings for training of the AI models. These vast pool of information are the very foundation and backbone of functioning of these AI models. However, much of this data is copyrighted and used without consent of the owner. The question as to the legality of such actions gives rise to uncertainty with respect to Copyright Law which was created in the analog era and has no explicit provisions for AI in most of the countries especially India. This technological  leap gives rise to the question whether training AI on copyrighted material would lead to infringement of copyright law. This study highlights the ambiguities in the current legal framework and the urgent need for legal reforms. This study would highlight the challenges posed by training of AI models under the Copyright Act 1957, as the conventional copyright law was developed to regulate the expressive exploitation of creative work. Whereas, training of AI models involves computational analysis of copyrighted work rather than expressive exploitation of the work such as adaptation, distribution, communication and reproduction. As AI works on large datasets, including the copyrighted material, it’s also imperative to find a striking balance between protecting creators ’rights and encouraging innovation and creativity. The study also aims to specifically investigate whether the works fed to the algorithms for statistical learning and training rather than for expressive use falls within the scope of section 14 of copyright act within existing exclusive right or if it can be categorised as a new category of use such as non-expressive computational use. The study aims to explore the legal challenges posed by technological advancements and propose a revised conceptual framework for addressing training of AI within copyright law in India while balancing the interest of the creators.