AI Training: Hybrid Model of Statutory Licensing on the Anvil

A hybrid model of statutory licensing with revenue-based royalties for the use of copyrighted works in AI training is on the anvil.

This proposal forms part of a DPIIT panel’s paper released on December 8, 2025. The panel was constituted earlier this year to examine the intersection of generative AI and copyright.

While this part of the working paper addresses issues relating to the use of copyrighted content in training AI systems, the second part, which will follow, will cover the copyright status of AI-generated outputs.

Stakeholder feedback is invited within 30 days.

Here’s an overview of the proposed framework:

• A mandatory blanket licence for copyright-protected works is envisaged, with no option for copyright holders to withhold works.
• There must be lawful access to the works. AI Developers cannot rely on the mandatory licensing framework to bypass technological protection measures or to gain unauthorised access to works behind paywalls without making the necessary payment.
• A centralised collective entity, comprising copyright societies and collective management organisations as its members (one member for each class of works) and designated by the central government, to collect payments from AI developers.
• A government-appointed rate-setting committee will fix royalty rates, which will be reviewed every three years.
• A certain percentage of the revenue generated from AI systems trained on copyrighted content will be payable as royalties. It follows that payment will become due upon the commercialisation of the AI system. Upon commercialisation, the royalty amount must be paid on an annual basis. The payment obligation will apply retroactively.
• Rates may be challenged before the court and will be subject to judicial review.
• AI developers must submit a disclosure form to the collective entity with details including category, source and nature of data used.
• Royalties received by the collective entity will be distributed to each member as per the share of each category of work disclosed in the form. Each member will then distribute royalties to copyright owners who have registered their works on an online system established by members. Only registrants will be eligible to receive a share of the royalties.
• The initial burden of proof to establish compliance in an infringement claim will be on AI developers.

The proposed model remains at a consultative stage, and the final framework will take form only after further discussions and stakeholder inputs. How these proposals evolve will depend on the responses received in the coming weeks.