Opinion India is trying to fix the copyright issue with AI training. It’s doing it wrong
India can adopt a framework that encourages innovation and become an AI powerhouse — or cling to relics of the past and watch that opportunity slip elsewhere
A blanket licensing regime would raise the cost of accessing generative AI, as the fees would be passed on to users, which would negatively impact these creators — and indeed all users of generative AI 
Written by Meghna Bal
On December 8, the Generative Artificial Intelligence and Copyright Committee, formed within the Department for Promotion of Industry and Internal Trade (DPIIT), released a paper that lays out the committee’s solution to the debate surrounding AI training data and copyright. Some rights holders — like news publishers, comedians, and authors — argue that training AI on their works without compensation is unlawful. AI companies counter that training does not involve reproducing or memorising protected works, but merely learning unprotected patterns and statistical relationships. They further argue that such use is transformative, a recognised defense in India and abroad. The clash between these camps has already sparked lawsuits around the world, including in India.
Seeking to quell this discord, the Committee proposes a blanket mandatory license covering all copyrighted works. Under this scheme, every AI model trained on copyrighted material would pay a flat rate based on a percentage of global revenue. The license would apply retroactively, meaning that existing, revenue-earning AI models trained on copyrighted content must begin paying. While a government-appointed committee would decide the rates, the body responsible for collecting and disbursing these fees would be a consortium of collective management organisations called the Copyright Royalties Collective for AI Training.
This structure is marketed as a win for all: AI developers gain legal certainty and “automatic access to training data”, and creators get guaranteed income. Yet these claims collapse under scrutiny.
First, the solution centralises power in Collective Management Organisations (CMOs) — institutions that have historically been at the root of many of India’s copyright troubles. CMOs were designed to make licensing efficient, as it would be impossible and costly for anyone wanting to use creative works to track down different artists and execute contracts with them. They were also meant to serve as a vehicle for strengthening creators’ bargaining power.
However, in the past, Indian CMOs have extorted users of works for large sums of money without being transparent about the rights they actually held, and they have colluded to fleece artists of ringtone royalties. Though the Copyright Act was amended in 2012 to rein in these abuses, CMOs that were no longer statutorily empowered to file claims on behalf of creators continued to do so. This checkered history begs the question: Why consolidate so much authority in institutions whose track record has continually undermined public trust? Entrusting these entities with administering a vast, mandatory AI licensing regime risks replicating past mismanagement on a far larger scale.
Second, CMOs represent only a fraction of India’s creative landscape. Vast swathes of artists — particularly emerging creators, independent authors, practitioners of traditional and folk arts, and entire categories of works for which CMOs do not exist — are not represented by them. For such artists, the Committee suggests the creation of a welfare fund to invest in the sector. Funds like these sound good in theory but often falter in execution. For instance, a recent report indicated that a Labour Welfare Fund under the Delhi Welfare Board held ₹5,200 crore in unspent funds. Further, it is unclear how allocations would be determined for smaller and lesser-known artists. The paper suggests that royalty distribution would prioritise the most popular works, raising serious concerns about whether lesser-known creators would see any meaningful benefit.
Third, the report presents the dynamic between AI companies and creators as a zero-sum game, ignoring the fact that many in the creative community have embraced AI systems to bring down production costs. A report by Classplus found that 92 per cent of Indian creators surveyed use generative AI. A blanket licensing regime would raise the cost of accessing generative AI, as the fees would be passed on to users, which would negatively impact these creators — and indeed all users of generative AI.
Fourth, the report’s emphasis on global revenue punishes scale and multinational presence for Indian AI start-ups, while also making it unclear whether they would be exempt from royalty requirements if their business were entirely in India. Making a firm liable for an additional fee once it goes abroad rests on the erroneous presupposition that all start-ups are formed to serve the domestic market first. There may be companies incorporated here but looking to serve needs abroad. Why should they be treated differently from those operating exclusively within the domestic market — even if they are the same size or perhaps smaller? It is akin to charging a tariff on the export of our own services.
The AI Copyright Committee’s blanket licensing proposal casts an unfortunate pall on the upcoming AI summit, where India is playing host while also seeking to position itself as a leader in AI innovation. Our conversations with Indian entrepreneurs make it clear that they cannot afford licensing costs or the expenses associated with hiring legal teams to manage additional compliance burdens. Faced with these costs, they will either refrain from entering the space or move abroad to jurisdictions with friendlier copyright laws. India now faces a choice: Adopt a framework that encourages innovation and become an AI powerhouse — or cling to relics of the past and watch that opportunity slip elsewhere.
The writer is director, Esya Centre, a tech-policy focused think tank based in New Delhi
Share
