India's government has proposed a framework for artificial intelligence training designed to keep data under the control of its owners, bringing AI models to secure, encrypted data stores rather than moving the data itself. The Data Empowerment and Protection Architecture (DEPA), outlined by the Ministry of Electronics and Information Technology (MeitY), would allow hospitals, banks and government agencies to share anonymised and encrypted data with a Data Exchange Platform, where model-makers could access it under conditions set by the providers.
Professor Ajay Kumar Sood, Principal Scientific Adviser to the Government of India, described the principle as "the model travels to the data, rather than the data travelling to the model." Under the plan, entities called Training Data Providers (TDPs) would share data with a Data Exchange Platform (DEP), where it would be stored in "hardware-based Secure Data Environments" that keep data encrypted even during processing. Training Data Consumers (TDCs) would access the data within these environments to train their models.
The framework specifies that differential privacy techniques would be applied to prevent models from learning individual-specific information, and all activities inside the secure environments would be logged for audit. India already operates several data exchanges, but it is unclear how much work remains to make DEPA operational.
MeitY envisions the architecture unlocking applications across several sectors. In healthcare, the framework says DEPA could “unlock currently fragmented health data to power breakthroughs in AI-driven diagnostics, drug discovery, and public health management.” In financial services, it could enable secure sharing of diverse financial data, potentially extending formal credit to 400 million underserved individuals and MSMEs. In education, the document says DEPA could overcome data interoperability challenges to enable AI-powered personalised learning for millions and create portable skill credentials for a future-ready workforce of its 250 million students in a $29 billion EdTech sector.
The document describes DEPA as India's contribution to global science diplomacy and also mentions the country's digital public infrastructure projects, such as the Unified Payments Interface.
Analysis
Why This Matters
- If implemented, DEPA could set a precedent for how governments regulate AI training data, balancing data localisation with model development.
- The framework addresses a core tension in the AI industry: the need for large datasets versus privacy and data ownership concerns.
- India's use of hardware-based secure environments for model training could influence other countries seeking data sovereignty.
Background
India has long pursued digital public infrastructure as a tool for economic inclusion and strategic autonomy. Its Unified Payments Interface and the Bharat Taxi ride-sharing platform are examples of government-backed systems designed to prevent Big Tech dominance. The DEPA proposal extends this philosophy to AI training. India's government has previously expressed concern that offshore model-makers will not address the country's specific needs or ensure local data owners remain in control.
What to Watch
- Whether the government proceeds with legislation or regulation to mandate DEPA for certain types of data sharing.
- How quickly existing data exchanges can be upgraded to meet DEPA's technical requirements.
- Adoption by healthcare and financial institutions as TDPs, and interest from AI developers as TDCs.
Checked
12 claims in this story checked against 1 source: 12 supported.
Show each claim
- SupportedIndia is floating an AI training plan in which the model travels to the data rather than the data travelling to the model.www.theregister.com - Articles: “As described by Professor Ajay Kumar Sood, Principal Scientific Adviser to the Government of India, DEPA means “the model travels to the data, rather than the data travelling to the model.””
- SupportedThe proposed DEPA framework would let AI developers train on encrypted, anonymised data without moving it.www.theregister.com - Articles: “The scheme envisions “Training Data Providers” (TDPs) – examples of which include hospitals, banks, and government agencies – sharing anonymized and encrypted data with a “Data Exchange Platform” (DEP). Entities called “Training Data Consumers” (TDCs) access the data stored at the DEPs to train their models.”
- SupportedIndia's government proposed DEPA to keep data owners in control, bringing models to secure data rather than moving the data itself.www.theregister.com - Articles: “India’s motive is to create AI models trained on local data, because it thinks offshore model-makers won’t be motivated to address the nation’s issues or to ensure owners of data stay in control.”
- SupportedProfessor Ajay Kumar Sood described the principle as “the model travels to the data, rather than the data travelling to the model.”www.theregister.com - Articles: “As described by Professor Ajay Kumar Sood, Principal Scientific Adviser to the Government of India, DEPA means “the model travels to the data, rather than the data travelling to the model.””
- SupportedTraining Data Providers would share data with a Data Exchange Platform, where Training Data Consumers could access it under rules set by providers.www.theregister.com - Articles: “They also enforce rules set by TDPs that describe how TDCs can use data.”
- SupportedData in the Secure Data Environments would remain encrypted during processing.www.theregister.com - Articles: “DEPs will operate what MeitY calls “hardware-based Secure Data Environments” in which “data remains encrypted and isolated even during processing.””
- SupportedThe framework says differential privacy would be used to prevent models from learning individual-specific information.www.theregister.com - Articles: ““Differential privacy techniques are systematically applied to prevent the model from learning or revealing individual specific information,””
- SupportedActivities inside the secure environments would be logged for audit.www.theregister.com - Articles: ““All activities inside the secure data environments are immutably logged for audit purposes, ensuring that the process is both secure and transparent.””
- SupportedIt is unclear how much work remains to make DEPA operational.www.theregister.com - Articles: “India’s government already operates several data exchanges, but it is unclear how much work remains to make DEPA a reality.”
- SupportedMeitY's framework says DEPA could use fragmented health data to support AI-driven diagnostics, drug discovery and public health management.www.theregister.com - Articles: “Once DEPA is working, MeitY envisions model-makers will use it to do things like “unlock currently fragmented health data to power breakthroughs in AI-driven diagnostics, drug discovery, and public health management.””
- SupportedThe framework says DEPA could potentially extend formal credit to 400 million underserved individuals and MSMEs.www.theregister.com - Articles: ““DEPA can enable secure sharing of diverse financial data, potentially extending formal credit to 400 million underserved individuals and MSMEs and fueling next-generation fintech innovation.””
- SupportedThe education section refers to 250 million students and a $29 billion EdTech sector.www.theregister.com - Articles: ““Similarly, in Education, DEPA can overcome data interoperability challenges to enable AI-powered personalized learning for millions and create portable skill credentials essential for a future-ready workforce of its 250 million students in a $29 billion EdTech sector,””
