Whalesbook
India Pushes for AI Copyright Royalties, Sparks Tech Sector Resistance
New Delhi – India's government is moving forward with a proposal to mandate royalty payments from artificial intelligence firms for data used to train their sophisticated models. The Department for Promotion of Industry and Internal Trade (DPIIT) aims to establish a framework to compensate creators whose works underpin AI development. This initiative, however, has encountered significant opposition from major technology companies.
Industry Pushback
Big Tech firms, represented by industry body Nasscom, argue that the DPIIT's plan, announced on December 8, is fundamentally unworkable. A dissent note published by Nasscom on December 17 highlights difficulties in separating commercial AI applications from early-stage, non-profit research projects. Such a distinction is crucial for determining royalty obligations.
The note also raises concerns about the fairness and governance of any proposed central body or collective management organization meant to distribute royalties. There's worry that niche and emerging works might not be adequately recognized or compensated under a standardized system.
Government's Rationale
Despite industry objections, the government maintains that regulating copyrighted data usage in AI is essential to foster creativity and prevent plagiarism. S. Krishnan, secretary at the Ministry of Electronics and Information Technology (Meity), emphasized the need to balance intellectual property rights with technological advancement.
"Concerns of plagiarism have existed for a long time, and copyright is recognized as a form of intellectual property in order to encourage creativity," Krishnan stated in an interview. He acknowledged the debate has two sides but affirmed that regulation is necessary in the AI age.
Technical and Legal Hurdles
Technology companies point to substantial legal and technical challenges. An internal note from a major tech group argues that the proposal unfairly shifts the burden of proof onto AI developers to demonstrate non-infringement, a task deemed technically infeasible for probabilistic generative AI models. The possibility of retrospective royalty payments for already-used data is also cited as particularly onerous.
Lawyers and industry experts echo these sentiments. Shailendra Bhandare, partner at Khaitan & Co., noted the complexity of long-term implementation and potential issues with mandatory licensing, which could override copyright holders' wishes. Global precedents, such as models in Japan and Singapore that permit data collection, are being observed.
Path Forward
Meity officials confirmed that the DPIIT paper is a discussion document intended to stimulate debate, not a finalized rule. The government is committed to further consultations to create a fair compensatory model and prevent future litigation, while regulating AI from a "harm's lens."
"We're not averse to regulating AI," Krishnan reiterated. "The matter of copyrighted content and its usage in AI is sub judice across the world." He pointed to existing court verdicts and private contracts between large AI companies and media organizations as precedents, suggesting a path toward collective bargaining and uniform agreements.
