The underlying issue — familiar to students of generative AI — of who owns the data, and who should profit from it, has now taken a new urgency with a government-appointed panel recommending mandatory blanket licensing and royalty. This model would force AI giants such as Google (Gemini), OpenAI(ChatGPT), and Meta to compensate content creators with a portion of their earnings for utilizing copyrighted works to train their models. This step, and especially the model advocated by the Indian government panel, represents a frontal challenge to the “fair use” doctrine that is now in favor among Big Tech companies like those based in nations such as the United States.
At its core, the problem here is economic asymmetry AI companies like OpenAI have built an enormous business successfully ingesting as much human-created content as possible — from news articles and photographs to books and music — because they render vast commercial value that is largely unshared with original creators. The E.U. panel recommended such practices because they “erode to a certain extent the economic interests of creators and artists, producers and content providers while benefiting internet platforms.”
India’s Hybrid Model: A Blueprint for the World?
The plan of the committee established by India’s Department for Promotion of Industry and Internal Trade (DPIIT) proposes a new “hybrid model” compensation that abandons two commonly seen regulatory models:
US “Fair Use” Model: AI companies claim that scraping data that is publicly available to train their systems is a form of “fair use”, for which they need not pay. The panel rejects that, pointing out that it would inherently put creators at a disadvantage.
EU ‘Opt-Out’ Model: The European Union also enables creators to “opt out” of training permission. The panel judged that this was not a solution, saying it would impose an unrealistic burden on creators to constantly monitor their work in gigantic, mysterious AI databases.
Instead, what the panel has recommended is a mandatory omnibus licence on which it wishes to superimpose a statutory right to remuneration and have it administered through a single body:
Mandatory access: AI developers would receive de facto and lawful guarantees of having access to all copyrighted content for training, leading to lower transaction costs as well as no individual negotiations.
Central Royalty Pool: A centralised non-profit organisation appointed by the government would receive royalty payments from AI developers.
Proposed Statutory Royalty Rate: A government commission would determine the royalty rate, but it could be calculated as a percentage of the AI company’s gross annual revenue for commercialized AI systems. This same rate will be used as retroactive to any successful commercial models that has already been trained on the content.
Everyone Pays, Everyone Gets Paid: These royalties would be paid not just to members of existing copyright organizations but also of other non-members, meaning indie labels or even artists would see some cash — so long as they register their works.
This unique-to-one-nation, mandatory-payment mechanism represents a break with contemporary global trends and situates India as a 21st-century outlier but also possibly ahead of the curve in protecting creator rights in the AI economy.
The Resistance and The Future of Fair Use
The technology industry has already vehemently objected to the proposal. Nasscom, a powerful tech industry body that represents Google and Microsoft, openly dissented, contending that any mandatory fee is akin to a “tax or levy on innovation” that would stymie development and unequally impact smaller startups. They argue in favor of a text-and-data-mining exception with an opt-out system — the very thing that the panel rejected.
The stakes are enormous. AI companies are already facing major lawsuits around the world, such as a precedent-setting suit between OpenAI and Indian news agency ANI over its use of unlicensed news content to train. If India adopts it, the global AI players will have to rework their business models for one of the world’s biggest and fastest-growing digital markets.
The panel’s recommendations are the subject of public and industry comment for 30 days. The next generation of generative AI must choose between two different futures: one in which data is free, and bereft of value for the people who supply it — this is the world that GPT-2’s creators envision — or a world where we acknowledge that miles logged on thought-roaders’ pacemakers are as valuable as their ideas, and create systems to ensure that all share in the economic benefits unlocked by automating “mental drudgery.”
