The rapid strides being made by LLMs in terms of quality, depth of information and sophistication of reasoning, are fuelled by two major factors. The first is iterative advancements in applied machine learning techniques, which continuously improve the performance of LLMs. The second is growing access to text, data and multimedia used in training these models. AI firms have argued that information on the Internet should be freely usable for such training, even if reproduction and similar acts of syndication by any non-AI entity would be subject to a licence fee and consent by the content producer. This has led to a fierce debate that pits the interests of AI hyperscalers and developers against content producers (news, entertainment and book publishing, to name a few). In this light, the Department for Promotion of Industry and Internal Trade’s working paper on AI and Copyright Issues’ proposal is a welcome step towards a solution where content providers are remunerated, without having a system that could put India’s AI ecosystem at a disadvantage.
The paper’s solution boils down to a mandatory licensing framework, which would allow AI data scraping tools to limitlessly scrape public information from the Internet, but would ensure that a non-profit copyright society-like body would collect payments from AI developers on the basis of revenue accrued to them through the benefits of training their models on Indian content producers’ data. The reasoning is partly driven by practical shrewdness on the drafting committee’s part, which rightly reasons that opting out of data scraping, and enforcing this when AI developers ignore their wishes, is not practical for every content producer. The other driver is a belief in data processing as a right for AI models, seeing as they do not usually reproduce content they are trained on, but synthesise fresh outputs. There are issues with how the royalty amount would be decided — small publishers who put in effort and investment in their work may chafe at receiving as much remuneration a piece as a big media house putting out hundreds of articles a day. But a system for remuneration is urgent, even if flawed. Lawsuits pitting AI firms against the publishers of the content they train models on are ongoing, and no uniform judicial thinking has materialised on the subject. The committee correctly notes that waiting for the outcome of any given litigation is unwise. The white paper — and a dissent by tech industry representatives — is a promising gesture towards a collaborative framework. The government needs to support this. A flawed system can be deliberated in the courts; the absence of one is a gift to AI firms in their ongoing disruption of the media and net landscape.
Published – December 11, 2025 12:20 am IST
