MeitY Launches Handbook for Safe Data Sharing in AI Training

The Ministry of Electronics and Information Technology (MeitY) launched the Handbook on Data Empowerment and Protection Architecture (DEPA) for AI Training on 31 October 2026. MeitY Secretary S. Krishnan unveiled it at a roundtable in Bengaluru. The iSPIRT Foundation supported the initiative’s development.

Framework for controlled data sharing

The handbook sets out a techno-legal framework for institutions to collaborate on AI training data while retaining control over their information. It aims to protect privacy, ownership, confidentiality, intellectual property and institutional control. Its safeguards combine contractual permissions, controlled computing environments and privacy-enhancing measures.

Data provenance and accountability

The framework uses data provenance mechanisms to track data origin, permissions and use throughout AI training. Principal Scientific Adviser to the Government of India Ajay Kumar Sood described its approach as enabling “the model to travel to the data, rather than the data travelling to the model”. The initiative is intended to make data sharing more secure and transparent and strengthen India’s AI ecosystem.

Important Facts for Exams

• Lineage and provenance: In data systems, lineage usually follows datasets and transformations across systems, whereas provenance focuses on how particular data items were derived and attributed.
• Provenance categories: Classical database research distinguishes why, where and how provenance, which can support query debugging, confidence estimation and annotation propagation.
• Consent-manager intermediary: DEPA's consent-manager model uses interoperable intermediaries to let individuals manage data-sharing permissions across otherwise separate data providers and users.
• Consent-artefact fields: A DEPA consent artefact can specify the data requested, recipient, purpose, duration and frequency, turning permission into a machine-readable, auditable instruction.