MeitY Floats New Privacy-Led Framework To Unlock Data For Training AI Models

The IT ministry (MeitY) has unveiled a consent-based framework called the Data Empowerment and Protection Architecture (DEPA) for AI training. This framework aims to enable startups, businesses, and academic institutions to access large-scale, India-specific datasets for AI model training while safeguarding citizens' privacy and ensuring compliance with the Data Protection Act.

The proposed DEPA for AI training also introduces a techno-legal architecture that allows organizations to collaborate on AI model training without directly transferring underlying datasets. This framework builds upon India's existing DEPA framework, which supports consent-based data sharing via initiatives like Account Aggregator.

The new framework includes several technical pillars: Confidential Clean Rooms, Differential Privacy, Electronic Contracts, and Zero-Trust Security. The framework also outlines key stakeholders, including Training Data Providers (TDPs), Training Data Consumers (TDCs), and Data Collectives or Data Collabs. TDPs register their anonymized and encrypted datasets on a data exchange platform, which are then tagged with metadata and cryptographic identifiers.

TDCs can request datasets for their specific use cases, and an electronic contract system verifies their requests against data-sharing policies and governance rules before granting access. Trained models must undergo independent checks for bias, fairness, security, safety, and explainability before pilot deployment and scaling. Post-deployment, oversight continues through continuous monitoring, periodic reviews, and traceability of model outputs back to the data and contracts used in their training.