Overview
- An eight-member DPIIT-led committee recommends a hybrid blanket licence letting AI developers use any lawfully accessed copyrighted works for training with no opt-out by rightsholders.
- Royalties would flow through a proposed Copyright Royalties Collective for AI Training (CRCAT), with rates set by a government-appointed panel and subject to judicial review.
- The paper contemplates flat, revenue-linked tariffs and requires dataset summaries to apportion payouts across classes of works, with obligations applying retroactively to commercialised models.
- Developers must show lawful access and cannot bypass paywalls or protection measures, while creators would claim from a central pool rather than negotiate individual licences.
- Nasscom and other industry groups formally dissent and urge a text-and-data-mining exception, as the proposal diverges from U.S. fair-use positions and lands alongside ongoing cases such as ANI v OpenAI in the Delhi High Court.