India Mandates AI Royalties for Creators in Bold Training Data Proposal

A government panel in India has proposed requiring AI companies to pay royalties to content creators whose works are used for training models, establishing a mandatory blanket license system. Led by the Department for Promotion of Industry and Internal Trade (DPIIT), the framework titled 'One Nation, One License, One Payment' rejects voluntary deals in favor of government-set rates collected by a new Copyright Royalties Collective for AI Training (CRCAT). This targets musicians, authors, news publishers, and others, ensuring equitable compensation through proportional distribution based on data usage categories like text or images.

The proposal diverges sharply from U.S. fair use doctrines embraced by OpenAI and Google, mandating lawful access and retroactive payments for past training on Indian content. Royalty rates, fixed by a government committee of experts and reviewed every three years, would apply as a revenue percentage, with AI firms submitting detailed dataset summaries for transparency. Challenges to rates allow judicial review, while the hybrid model prohibits bypassing paywalls or protections.

Industry voices raise concerns over government-fixed pricing stifling innovation and bargaining, with groups like Nasscom warning of barriers for startups. Proponents argue it levels the playing field for small creators against Big Tech's negotiating power, potentially making India the first nation with statutory AI training royalties. The working paper invites 30 days of stakeholder feedback before finalization.

This move aligns with global copyright debates, pressuring AI giants reliant on public data in India's vast market. Enterprises face higher costs for model development, prompting shifts toward licensed datasets or local alternatives. As AI ethics evolve, India's framework could inspire similar protections worldwide, balancing innovation with creator rights