The Pulse
India Launches DEPA Handbook for Trusted AI Training
India’s Ministry of Electronics and Information Technology launched a handbook for using its Data Empowerment and Protection Architecture in AI training. The framework combines contractual permissions, controlled computing environments and
AI.info Team · Written with AI · fact-checked ·
India’s MeitY launches the DEPA training handbook
India’s Ministry of Electronics and Information Technology launched a handbook for trusted AI training on October 31, 2026, at a roundtable in Bengaluru. The Data Empowerment and Protection Architecture, or DEPA, framework is intended to let organizations collaborate on training data without giving up control of the information they hold. MeitY Secretary S. Krishnan unveiled the handbook with support from the iSPIRT Foundation.
The announcement addresses a persistent institutional barrier: organizations may hold large, useful datasets but hesitate to share them because of privacy, confidentiality, intellectual-property and control concerns. DEPA proposes a structure for purpose-bound collaboration, in which permissions define how data can be used and institutions retain authority over their data. The government describes it as a “techno-legal framework,” combining contractual permissions with controlled computing environments and privacy-enhancing safeguards.
Permissions and provenance shape how data is used
Data provenance is another part of the design. The framework aims to trace where data came from, what permissions apply and how the data is used during training. That record is intended to support accountability and transparency, though the announcement does not spell out the technical standards or auditing process that would make those aims measurable.
For organizations that might otherwise decline to share sensitive records, the approach offers a way to participate under defined conditions rather than surrendering custody. Hospitals, banks, research institutions and other data holders could be relevant participants, but MeitY’s announcement does not identify a formal eligibility list or detail how organizations would join. It also does not specify which computing environments would qualify or how compliance would be checked.
“The model travels to the data”
MeitY Secretary S. Krishnan framed the initiative as a way to expand access while protecting the interests of data holders. Krishnan, Secretary, Ministry of Electronics and Information Technology said: “India’s AI ambitions must be matched by the ability to use data responsibly and securely.”
Ajay Kumar Sood, Principal Scientific Adviser to the Government of India, described the operating principle in direct terms: Sood, Principal Scientific Adviser to the Government of India said, “Under DEPA, the model travels to the data, rather than the data travelling to the model.” In practice, the stated objective is to support collaboration among data holders and AI developers while keeping data use within agreed permissions. Sood also pointed to multi-party and cross-border collaboration as potential applications, linking the framework to joint scientific work.
Implementation will determine the framework’s reach
The launch brought together representatives from government, industry, technology, academia, research and media. The roundtable included a presentation on the training framework by Hari Subramanian, alongside perspectives from representatives of Emergent AI, Sarvam, HDFC Bank and other organizations. The event indicates interest across sectors where data is valuable but often difficult to share; the announcement does not report deployments, adoption figures or completed training projects under the new handbook.
That leaves practical questions for implementation. The announcement does not set out the terms organizations will use to grant or receive permissions, the safeguards against misuse of trained models, or how disputes and failures would be handled. Nor does it describe how the framework will be tested against the different legal and technical requirements of sectors such as banking and healthcare. Those details will determine whether DEPA becomes a usable route for shared AI training or remains a statement of principles.