India's ambition to become an AI powerhouse hinges on one critical resource: data. The Ministry of Electronics and Information Technology (MeitY) has just unveiled a plan to unlock this resource without getting tangled in fresh legislation.
A Framework, Not a Law
Instead of
drafting a new, rigid law for a rapidly evolving technology, MeitY has launched a handbook for the Data Empowerment and Protection Architecture (DEPA). This is a 'techno-legal' framework, meaning it combines existing legal permissions with technological safeguards to allow companies, hospitals, banks, and government agencies to share data for training AI models. The goal is to address the persistent problem of AI startups and researchers in India lacking access to large, high-quality, India-specific datasets. This move is part of a broader strategy to foster innovation by creating adaptable guidelines rather than laws that could quickly become outdated. India has consistently signalled its intent to regulate AI through existing statutes where possible, avoiding the path of comprehensive, standalone AI legislation like the EU's AI Act.
How 'The Model Travels to the Data'
The core concept of the DEPA framework is a paradigm shift in data sharing. As Principal Scientific Adviser Ajay Kumar Sood explained, the plan is for 'the model to travel to the data, rather than the data travelling to the model'. In practice, this means data holders, termed 'Training Data Providers', won't send their raw datasets to AI developers. Instead, developers, or 'Training Data Consumers', will train their algorithms within secure, controlled computing environments managed by intermediaries sometimes called 'Data Collabs'. This allows multiple institutions to contribute to an AI training exercise without ever losing control of their underlying sensitive or commercially valuable information, a major concern that has historically limited data sharing.
The Safeguards in Place
The DEPA framework proposes a multi-layered approach to security and privacy. It combines contractual agreements that limit data use for specific purposes with technological measures. Key safeguards include 'Confidential Clean Rooms' or 'Secure Data Environments', which are isolated digital spaces where data can be processed without being exposed. Another technique is 'differential privacy', which involves adding statistical noise to datasets to prevent the identification of any single individual. The framework also mandates 'data provenance' mechanisms, creating a transparent and auditable trail of where data came from, what permissions were granted, and how it was used throughout the AI training process.
Navigating Privacy and Consent
One of the most significant aspects of the DEPA handbook is its approach to consent. The document notes that obtaining individual consent from millions of people is often impractical for large-scale AI training. Instead, it aims to protect user interests through the robust technological and contractual safeguards built into the system. This approach appears to leverage exemptions within the Digital Personal Data Protection (DPDP) Act, 2023, which allows for the processing of personal data for research and statistical purposes under certain conditions. By de-identifying data and ensuring it is not used to make decisions about specific individuals, the framework aims to comply with the spirit and letter of existing data protection laws.
Fueling India's AI Ambitions
This new framework is a cornerstone of the broader IndiaAI Mission, a government initiative to bolster the country's AI ecosystem with significant financial backing. By creating a trusted mechanism for data sharing, MeitY hopes to unlock innovation in critical sectors. For example, pooled health data from sources like the Nikshay portal for tuberculosis could build early-warning systems for disease outbreaks. In finance, aggregated and de-identified data could help train models to extend credit to millions of underserved individuals and small businesses. For startups and researchers who have been data-starved, DEPA promises access to the fuel they need to build and scale AI solutions tailored to India's unique challenges and diverse population.
