On October 29, 2026, MeitY released a framework called “Data Empowerment for AI Training” (DEPA-AIT). This document is an initiave for strengthening the development of indigenous foundation models making available training data in a systematic manner. It poposes a national framework to make large and diverese Indian datasets available for AI model training. It is expecte that it will be protecting privacy, confidentiality, security and institutional interests.
At a time when we are discussing the DPDPA implementation, the entities holding the data are concerned about the consequences of data misuse both under ITA 2000 and DPDPA 2023. Hence they will be reluctant to share any of their data. This “Data Deadlock” is set to be broken by the adoption of this DEPA-AIT.
Knowing the influence of the Big Tech on MeitY through NASSCOM, we need to however flag the risk of foreign AI models stealing the data before the Indian AI players can have their hands on it.
A “Guardrail” to prevent this misuse has to be in place before any thoughts can be expressed on the proposal.
We need to recall that the CIC set up was enabled in India to protect the Indian banks but it only enabled TransUnion, a US company gain access to all the sensitive data that was available to CIBIL. A similar possibility exists in the DEPA-AIT framework and Chat GPT or Google or Caude may be the biggest beneficiaries.
We need to study the framework in more detail to understand what kind of gruard raisl are built to prevent such misuse.
A copy of the Framework is available here:
More discussions will follow.
Naavi
