Pillar 04

Sovereign AI

Data localization, sovereignty, and an India-based practitioner's view.

Sovereign AI usually gets argued as politics. For an enterprise architect it's a design constraint — concrete questions about where a model runs, who can reach the data it processes, and what happens to that data after inference. This pillar treats sovereignty as five buildable dimensions — infrastructure, data residency, model ownership, operational independence, and governance — and makes the case that small language models have turned it from aspiration into something you can actually ship. The real test isn't ideology; it's telling apart the dependencies you chose from the ones you merely inherited. The vantage here is a practitioner's — grounded in running these compliance reviews, not theorising about them.