India’s Department for Promotion of Industry and Internal Trade (DPIIT) has released a working paper on generative AI and copyright that could significantly reshape how AI models are trained on copyrighted content in the country. At its core, the paper backs a mandatory “blanket licence” for AI training, paired with a statutory royalty right for creators.
The document maps two big pressure points:
- Unlicensed use of copyrighted material as training data (inputs)
- Uncertainty over who owns or controls AI-generated outputs
It concludes that India’s Copyright Act lacks clear exceptions for text and data mining or AI training, creating legal risk for both rightsholders and developers.
To address this, the paper recommends a hybrid statutory licensing model wherein AI developers could train on any “lawfully accessed” content under a compulsory blanket licence, without negotiating a high number of individual permissions. In return, copyright owners get a statutory right to remuneration, with royalties triggered when AI systems are commercialised and rates set by an independent, government-designated body.
The model effectively creates a central organisation that would collect and distribute royalties, cutting transaction costs and offering legal certainty to both big tech and startups. The committee explicitly rejects a “zero-price licence” or broad free-use exception, warning that free access to all copyrighted works would undermine incentives for human creativity over time.
If implemented, India would be among the first countries to hardwire a compulsory AI training licence into copyright law, positioning itself as a potential reference point for IP norms with respect to AI.
DPIIT has opened a 30-day public consultation, inviting comments from creators, platforms, AI firms, and civil society. Key questions include whether denying rightsholders an opt‑out from training is consistent with international IP obligations, and how fairly royalties can be distributed across diverse creative sectors.
For India’s AI ecosystem, the paper locks in a state-mediated data and revenue layer. Considering how big AI firms like OpenAI and Meta have been able to access copyrighted material to train their models, sparking lawsuits and regulatory turmoil worldwide, this proposed hybrid licensing model could level the playing field and reduce regulatory friction. By granting startups equal access to lawfully sourced data via a mandatory blanket licence and avoiding the need for endless individual negotiations, this approach can also accelerate innovation while compensating rightsholders.
However, how this hybrid licensing system is designed in detail, including governance, transparency, and dispute resolution, will determine whether it becomes a catalyst for innovation or another friction point for India’s tech ecosystem.