MeitY’s DEPA training handbook pairs data permissions with controlled computing

MeitY unveiled its DEPA for AI Training handbook in Bengaluru on October 31. The ministry’s [announcement](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2330756&lang=1®=20) describes purpose‑bound collaboration combining contractual permissions, controlled computing and privacy safeguards. Provenance is intended to trace data’s origin, permissions and use throughout training. Principal Scientific Adviser Ajay Kumar Sood described the model travelling to the data, rather than data travelling to the model.

For a data custodian, that separates the permitted training purpose from the computing environment. Keeping data in place does not itself answer who may use it. The [release](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2330756&lang=1®=20) announces a framework; the full handbook and an operational deployment were not reviewed. For architecture review, the [Permission‑aware data](https://opentools.ai/resources/enterprise‑system‑design) reference discusses authority and data boundaries. It offers a checklist for builders, not DEPA documentation. Indian Railway 139 server room, August 2009; archival computing context, not a DEPA installation. Photo: [Indrajit Das](https://commons.wikimedia.org/wiki/User:Indrajitdas) / [Wikimedia Commons](https://commons.wikimedia.org/wiki/File%3A139_Server_Room_01.jpg), [CC BY‑SA 3.0](https://creativecommons.org/licenses/by‑sa/3.0). Resized and cropped under the same licence.