Practical AI Nepal organises around six strategic pillars. Together they describe what it actually takes for Nepal to use, govern, run, adapt, evaluate, and improve AI on its own terms — not as theory, but as institutional and community practice.
AI sovereignty is not isolation. It is capability and optionality — the ability to choose, govern, evaluate, run, adapt, and continue when external platforms change.
Governance is not a PDF. It is what changes how procurement, review, audit, and disclosure actually happen — the procedures and habits that hold an AI system accountable from intake to incident.
Sovereignty is the ability to choose appropriate tools, govern data and risk, evaluate model limits, run AI in the right infrastructure, and continue working when external platforms change. Capability and optionality, not isolation.
The capability to test and adapt open and closed models for Nepali language, local institutions, domain knowledge, and real-world tasks — and to decide, deliberately, which approach fits which problem.
The choice between a public API, a private cloud, a hosted open model, a local GPU, an on-premise system, or an on-device model determines who can see the data, who is accountable, what it costs, and how it can be audited.
Nepal's AI future should include students, schools, colleges, rural communities, mobile devices, local labs, and volunteers. Edge AI lets communities participate — evaluating, validating, testing — without centralising every dataset.
AI readiness requires more than tools. It requires literacy, trained executives, prepared teachers, capable journalists, upskilled developers, contributing students, and institutional habits that survive turnover.