RECONCILING GENERATIVE ARTIFICIAL INTELLIGENCE AND COPYRIGHT LAW: ANALYZING TRAINING DATASET INFRINGEMENT AND OUTPUT AUTHORSHIP IN INDIA

As Generative Artificial Intelligence (GenAI) applications rapidly evolve, the entire global IP landscape is changing, making it urgent to tackle the conflict between the classic doctrines of copyright and automated neural networks as a whole. The first legal question I have considered in this paper is whether the use of copyrighted works for creating AI-based training models is a misuse of such copyrights or does it fall under the ‘fair dealing’ provisions of Part VI of Indian Copyright Act, 1957 under Section 52(1), and secondly whether the autonomous production of AI under the use of such works fulfils the provisions of Section 2(d) of Indian Copyright Act, 1957. This study also reveals that the closed list model of ‘fair usage’ in India poses legal uncertainties as compared to the model of fair dealing the USA and the computer-generated work framework with their open lists in the UK. There are no existing research exceptions for commercial data ingestion and pure AI outputs do not have creative agency of humans. To break the deadlock, the article recommends a generic statutory reform of the copyright framework with a dual tier TDM exemption, transparency of the datasets and a statutory collective licensing framework for compensation for human creators.