The Department for Promotion of Industry and Internal Trade has come out with a working paper to frame rules for how AI models can use copyrighted works for training, and how the resulting value should be shared with human creators. Rishi Raj explains what exactly it recommends & the reactions it has triggered
l What is this DPIIT paper and why now?
THE DEPARTMENT FOR Promotion of Industry and Internal Trade (DPIIT) has released a working paper on Generative AI and copyright as a draft policy proposal, open for public comments. The trigger is that large AI models are being trained on massive quantities of text, images, music and video, much of which is protected by copyright, and courts abroad are seeing lawsuits alleging unauthorised use. India does not yet have clear rules on whether training AI on copyrighted works counts as infringement, or whether it falls under existing fair dealing exceptions, and the paper is meant to plug this gap as some cases have reached the courts here also.
l What problem is the government trying to solve
AT THE CORE is a tension between two objectives. On the one hand, the government wants to build an AI ecosystem and cannot realistically expect startups or even large firms to negotiate with every author, publisher or label whose work might end up in a training dataset. On the other hand, creators argue their work is being used to build commercially valuable systems without consent or compensation.
The paper’s aim is to create a framework where AI developers get lawful access to training data while creators receive a guaranteed payment, without litigation.
l What is the actual proposal?
THE PANEL HAS proposed a hybrid statutory licensing model summarised as One Nation, One Licence, One Payment. Under this, AI developers get a mandatory blanket licence to use all lawfully accessed copyrighted works for training models. Creators, in turn, get a statutory right to remuneration, with royalties payable only when AI systems are commercialised and start generating revenue. These royalties would be set by a designated authority, subject to judicial review, and collected and distributed through a central mechanism designed as a single window.
l Why not text-and-data mining exceptions?
SEVERAL JURISDICTIONS, ESPECIALLY in Asia and parts of Europe, have chosen the route of broad text-and-data-mining (TDM) exceptions that allow AI training without permission, at times with opt-out rights for rights holders. The DPIIT committee considers this option but ultimately discards it on two grounds. First, it would leave creators without any structured way to claim compensation. Second, in a country of India’s size and diversity of content industries, administering opt-outs and tracking who has excluded their works would introduce its own complexity without resolving the underlying fairness concerns around unpaid use.
l Concerns about the blanket nature of the licence
OBJECTIONS HAVE BEEN raised by some because under this model, creators do not have a meaningful right to refuse the use of their work for AI training. For many, this cuts against the grain of copyright as a property right based on individual consent and control. Legal experts maintain that such compulsion may raise Constitutional questions around property and due process. Industry bodies like Nasscom have also objected, stating that a mandatory, centralised licensing architecture, coupled with regulated royalties, risks burdening innovation and could become a de facto tax on AI development.
l Why the paper calls it a pragmatic approach?
THE COMMITTEE’S REASONING is rooted in AI’s scale and opacity. Modern models ingest billions of data points scraped from the open web and licensed sources, making it practically impossible to negotiate permissions on a case-by-case basis or to give each creator a veto. So, a purely consent-based system would either bring innovation to a halt or lead to an uneven playing field where only the biggest technology companies can clear the necessary rights.
l Effect on startups vs Big Tech
MUCH OF THE debate centres on whether the proposal helps or hurts smaller firms. A single-window statutory licence removes the need to sign hundreds of bilateral contracts, which ought to help startups. But there are compliance burdens, future royalty rates and the risk that a centralised system could be captured by larger interests. There is also concern that if royalties are set too high, global players may absorb the cost more easily than domestic ones, diluting the objective.
l What next
THE PAPER HAS been released as the first part of a broader exercise and stakeholders can offer feedback and comments over the next month.
