India’s AI ambitions require massive amounts of data, but privacy and security are major concerns. The Ministry of Electronics and Information Technology (MeitY) just released a new rulebook to solve this puzzle.
What is this New Framework?
MeitY has launched the 'Handbook on Data
Empowerment and Protection Architecture (DEPA) for AI Training'. This isn't a new law, but a detailed techno-legal guide for institutions like hospitals, banks, and government agencies to share their vast datasets with AI developers, researchers, and startups in a secure way. The core idea is to unlock valuable data to train artificial intelligence models without the data holders losing control or compromising privacy, intellectual property, or commercial confidentiality. This structured approach aims to break down the data silos that currently limit AI development in the country.
The 'Model Travels to the Data' Approach
A key principle of the DEPA framework is a concept described as 'the model travels to the data, rather than the data travelling to the model'. Instead of institutions sending their raw datasets to AI companies, the AI models are brought into a secure, controlled computing environment where the data is held. The model gets trained on the data within this secure bubble, and only the refined AI model leaves—the underlying data never does. This significantly reduces the risk of data leaks and misuse, addressing a major hesitation for organisations considering data sharing.
How is Privacy Protected?
The framework acknowledges that simply removing names and IDs is not enough to guarantee anonymity, as individuals can often be re-identified by combining different pieces of information. To counter this, DEPA proposes a multi-layered defence. It combines strong contractual agreements, the use of secure sandboxes for training, and advanced privacy-enhancing techniques like 'differential privacy', which adds statistical noise to data to make it impossible to isolate individual records. Furthermore, it mandates robust data provenance mechanisms, which create a clear audit trail of where data came from, who has permission to use it, and how it was used, ensuring transparency and accountability.
Who are the Key Players?
The DEPA handbook identifies three main roles in this new ecosystem. 'Training Data Providers' are the institutions holding the data, such as banks or hospitals. 'Training Data Consumers' are the startups, researchers, and companies looking to train their AI models. Acting as intermediaries are entities called 'Data Collabs', which will be responsible for pooling data from various providers, potentially for specific sectors like healthcare or finance, and facilitating the secure sharing process. This structure is designed to be managed by a proposed India Data Management Office (IDMO), which will set the rules and standards for this entire data-sharing ecosystem.
Why This is a Big Deal for India's AI Ambitions
For years, the Indian tech industry has faced a significant hurdle: a lack of access to large, high-quality, India-specific datasets needed to build relevant and effective AI solutions. While the country generates enormous amounts of data, it remains locked away due to valid privacy and security fears. This framework is a direct attempt to solve that problem. By creating a trusted and secure pathway for data collaboration, the government hopes to fuel a wave of innovation. It could empower Indian startups and researchers to create cutting-edge AI in fields like healthcare, finance, and agriculture, ultimately boosting the digital economy and strengthening India's position in the global AI landscape.
