India’s Data-First AI Plan Puts Control Before Model Scale

India wants AI models trained without exporting local data. Its Ministry of Electronics and Information Technology (MeitY) has proposed the “Data Empowerment and Protection Architecture (DEPA) for AI Training,” a framework designed to regulate information flows while keeping data owners in control.

Professor Ajay Kumar Sood, Principal Scientific Adviser to the Government of India, described the central idea in one clean line: “the model travels to the data, rather than the data travelling to the model.” That reverses the usual arrangement, where data gets gathered into systems built by model-makers and control becomes someone else’s problem.

India’s stated motive is practical. It wants AI trained on local data because offshore model-makers may not have the motivation to address the nation’s issues or ensure that data owners retain control.

How DEPA Would Keep Data in Place

DEPA envisions hospitals, banks, and government agencies acting as “Training Data Providers,” or TDPs. They would share anonymized and encrypted data with a “Data Exchange Platform,” while “Training Data Consumers,” or TDCs, would access data stored at those platforms to train their models.

The exchange platforms would operate what MeitY calls “hardware-based Secure Data Environments.” Inside those environments, “data remains encrypted and isolated even during processing,” so the data does not need to take a sightseeing trip through every model-builder’s infrastructure.

Each TDP would set rules describing how TDCs can use its data, and the DEPs would enforce those rules. MeitY’s framework says, “Differential privacy techniques are systematically applied to prevent the model from learning or revealing individual specific information,” while “All activities inside the secure data environments are immutably logged for audit purposes, ensuring that the process is both secure and transparent.”

That gives DEPA three control points: the data stays with a governed exchange, privacy protections apply during training, and the system records activity for audits. India’s government already operates several data exchanges, but it is unclear how much work remains before DEPA becomes a working system rather than a framework with excellent paperwork.

Health, Finance, Education and Diplomacy

MeitY sees DEPA as infrastructure for sectors where useful data is scattered across institutions. In health, it says the framework could “unlock currently fragmented health data to power breakthroughs in AI-driven diagnostics, drug discovery, and public health management.”

Finance is another target. India is home to literally thousands of banks, and MeitY says DEPA could enable secure sharing of diverse financial data, potentially extending formal credit to 400 million underserved individuals and MSMEs while fueling next-generation fintech innovation.

The education case is also built around fragmented systems. The framework says DEPA could overcome data interoperability challenges, support AI-powered personalized learning for millions, and create portable skill credentials for a future-ready workforce of 250 million students in a $29 billion EdTech sector.

MeitY calls adoption across these areas “the essential catalyst for realizing India’s vision of inclusive growth and responsible AI leadership on the global stage.” It also presents DEPA as “India’s contribution to global science diplomacy,” advancing efforts by the World Bank and the United Nations to shape a more inclusive and responsible global AI ecosystem.

The framework makes several mentions of digital public infrastructure, an idea pioneered by India through government-funded tools such as Unified Payments Interface payment systems and the Bharat Taxi zero-commission ride-sharing platform. India uses DPI to prevent Big Tech companies from dominating the local market and as a diplomatic tool by giving it to friendly nations.

AI Governance Is Becoming Operational

DEPA is one proposal in a wider set of attempts to place rules around AI and data systems. At the Global Data Ecosystem Conference 2026 in Shanghai, whose theme was “Data Intelligence Connecting the World,” the China Meteorological Administration introduced the Shanghai Node for International Exchange of Earth System Data and the Trusted Data Space for the Meteorological Industry.

Japan’s Digital Agency has also started a trial of an app that helps draft answers to lawmakers’ questions in parliament. The app uses the agency’s Gennai generative artificial intelligence platform, which is available to about 180,000 government workers, and helps officials understand the intent behind questions before they prepare responses.

When officials enter a lawmaker’s question, the app organizes and analyzes its key points, then generates multiple draft answers within minutes using past government answers and minutes of Diet deliberations. It can adjust writing styles, offer a “policy explanation” mode for clarifying policy direction, and provide a “political judgment” mode that reflects political considerations.

Officials can edit the drafts and use the app to predict follow-up questions. The Digital Agency plans to evaluate its effectiveness during the current session of the Diet, Japan’s parliament, with full-scale operation targeted from next year’s ordinary Diet session.

Digital transformation minister Toshiharu Furukawa said, “It’s very good that we can improve our work efficiency and create answers that contribute to deeper Diet deliberations,” adding, “I’d also like to hear feedback from other ministers.” The agency stressed that officials remain responsible for clarifying policy decisions, organizing the government’s views, and finalizing answers; decision-making will not be left to AI.

That distinction sits at the heart of India’s proposal too. DEPA is not promising that models deserve authority; it is proposing controlled access, traceable use, and human ownership of the rules. The model may travel to the data, but accountability still needs to know where it lives.

Stay connected via Google News

Follow us for the latest AI News