India floats DEPA AI plan: model travels to data, not data to model

The proposal comes as legal argument continues over how far large language model developers may feed the world's data into their systems without paying. India's stated 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 data owners stay in control.

Under the DEPA framework, Training Data Providers such as hospitals, banks and government agencies share anonymized and encrypted data with Data Exchange Platforms. Training Data Consumers then access that data inside hardware-based Secure Data Environments, where differential privacy techniques are applied and all activity is logged for audit. MeitY frames the scheme as India's contribution to global science diplomacy, citing efforts by the World Bank and the United Nations to shape a more inclusive global AI ecosystem, and it ties the plan to India's push for digital public infrastructure as a way to keep large technology companies from dominating the local market.