India's DPIIT committee has put forward a significant working paper addressing AI large language models' (LLMs) access to online content amid copyright concerns, releasing the draft in December 2025 and inviting public feedback for 30 days. The committee recommends that LLMs, exemplified by ChatGPT, be allowed free access to content freely available on the internet with no opt-out option for publishers, aiming for a clear legal framework that supports AI innovation while compensating content creators. To that end, the paper proposes creating a non-profit copyright society to collect royalties for the use of such content, encompassing both members and non-members of the society, akin to India’s compulsory licensing framework for radio stations. This model would impose a statutory fee payable to rights holders, reducing ongoing copyright litigation uncertainties exemplified by the Digital News Publishers Association’s lawsuits against OpenAI. Industrial dissent was voiced by Nasscom, which opposed mandatory royalties and favored freedom from paying for content scraping, arguing mandated fees would stifle innovation and that content owners should have an opt-out mechanism. The committee countered that smaller content creators lack the means to enforce opt-outs, underscoring the need for a blanket, enforceable system. The royalty distribution scheme is designed to consider web traffic metrics and publisher credibility, with formal appeal channels through judiciary systems to resolve disputes. The recommendation reflects an effort to balance rights protection and AI development pressures, although it may face challenges from both AI firms, concerned about additional costs when many are currently unprofitable, and from publishers seeking more valuable compensations for their content’s role in AI training.