India Signals 3-Year Upper Limit For New AI Copyright Law

India is preparing for a multi-year overhaul of its copyright framework for the age of artificial intelligence, with the government indicating that a final legal architecture has a three-year horizon to materialise.

Part 2 of the DPIIT committee’s working paper, which will address whether AI-generated works can receive copyright protection and who should be considered their author, will be released next.

In the first part of its paper, released on December 8 and clarified through an official statement on Thursday, the commerce ministry's Department for Promotion of Industry and Internal Trade proposed a new hybrid model that would allow AI developers to train on all lawfully accessed copyrighted content without seeking permission from creators. The government has opened a 30-day feedback window before moving toward drafting amendments.

Under the proposal, creators will not be able to opt out of having their works used in AI training, nor will AI developers need to negotiate with them individually. DPIIT said global experience showed that opt-out systems had low success rates and largely failed smaller creators, who lack the awareness, bargaining leverage and technical tools to prevent scraping even after opting out. Negotiated licencing models, it said, introduced prohibitive delays and high transaction costs that could freeze out startups.

Instead, the proposed system would grant AI companies a blanket licence to use all accessible content for training, with creators compensated only after the resulting AI tool is commercialised. No upfront fees would apply, making the model retroactive in effect, and all existing copyrighted works could be used for training immediately.

A new body, the Copyright Royalties Collective for AI Training, would collect royalties and distribute them through copyright societies and collective management organisations.

Both members and non-members would be eligible to receive payments as long as they register their works. Royalties for sectors without a functioning CMO would be held for three years; unclaimed amounts would then flow into a welfare fund for that sector.

The framework notably avoids EU-style transparency rules that require disclosure of datasets used to train models. DPIIT said such requirements could slow innovation, particularly for MSMEs. Paywalled content would remain off-limits, but everything else accessible lawfully, including public text, images, audio and video, could be used for training by default.

The committee examined global models from the US, UK, EU, Japan and Singapore but found each inadequate. Exceptions for text-and-data mining, voluntary licencing frameworks and blanket exemptions were seen as either too burdensome or too skewed to large industry players. India's model aims to guarantee data access for AI developers while offering creators a predictable revenue stream.

With Part 2 focused on authorship and copyrightability set to be out soon, DPIIT’s proposal marks one of the most ambitious attempts globally to reshape copyright law for India's AI era.