India's IT Industry Pushes Back Against Proposed AI Training Rules

Image Source: Internet

India's National Association of Software and Service Companies (Nasscom) has expressed strong opposition to a proposed hybrid model that would require AI companies to pay royalties for training on copyrighted content. The model, presented by the Department for Promotion of Industry and Internal Trade (DPIIT), has been met with resistance from big tech firms and industry experts, who argue that it would be difficult to implement and would make AI development expensive and legally risky. The proposed model would mandate a blanket licence for AI training on copyrighted content, with royalties paid into a central body called the Copyright Royalties Collective for AI Training (CRCAT). However, industry experts point out that general-purpose AI models like ChatGPT operate on thin margins and would struggle to sustain mandatory royalties. They also argue that the model would be difficult to enforce, as AI developers can further reduce traceability by increasing randomness or adding noise to models. Several industry executives and experts have come forward to express their concerns about the proposed model. Jameela Sahiba, an associate director at tech policy think tank The Dialogue, warned that the move would hurt India's startup ecosystem, as startups operate with limited capital and would struggle to meet the compliance requirements. Another industry executive alleged that the committee's setup played a huge role in the recommendation, with members having a vested interest in favouring publishers. The proposed model also contradicts the AI governance framework released by the Principal Scientific Advisor on November 5, which asked for a balanced approach that enables Text and Data Mining while protecting the rights of copyright holders. The IT ministry has backed the recommendations by the DPIIT committee, despite the framework not being legally binding. Nasscom has formally dissented to the proposal, citing technical, economic, and enforcement challenges. The trade body has suggested that India should allow Text and Data Mining for AI training as long as the content is accessed legally, and that creators should be able to block AI training through a machine-readable opt-out or contractual restrictions. The full 24-page written submission by Nasscom has been added as an annexure to the report, despite the DPIIT committee's initial plan to only include a summary.