AI Governance in South Asia
South Asia’s pragmatic AI governance path will only become credible when soft-law coordination is backed by enforceable controls, sovereign infrastructure choices, and real labour protections.
Review
South Asia is quietly converging on a distinctive AI governance posture: pragmatic, use-case-driven, and pro-adoption rather than pro-spectacle.
“AI Governance in South Asia” makes that case clearly across India, Bangladesh, Sri Lanka, Nepal, and Bhutan, where AI is being framed as development infrastructure rather than frontier competition. The focus is agriculture, health, language access, disaster response, and public services. That instinct is sound.
But convergence of intent is not convergence of power.
Most governance in the region still lives in guidelines, principles, and advisory bodies. Liability remains fuzzy. Enforcement is uneven. Revocation and redress are largely absent. Soft law is doing too much work where hard institutional capacity is missing.
A use-case-first approach accelerates deployment. It also accelerates dependency. Without control over data flows, compute allocation, and model adaptation, scaling AI for public value can also scale foreign reliance. Infrastructure sovereignty cannot remain an aspiration if AI is becoming critical public infrastructure.
Labour transitions are the clearest fault line. More than sixty percent of the workforce across these countries is informal. Yet AI governance largely assumes formal employment, formal contracts, and formal recourse. Reskilling without enforceable protections is not transition management. It is risk displacement.
The regional collaboration story is compelling, but fragile. Forums do not create governance. Control planes do. Trusted data corridors, shared compute, and language commons only work if someone owns authority, accountability, and failure modes, and treats revocation, escalation, or machine-enforceable authority as first-class primitives.
South Asia has an opportunity to shape a third path in global AI governance, neither precautionary paralysis nor market fundamentalism. But that path will only be credible if governance moves from principles to execution, from coordination to control, and from aspiration to enforceability.
Otherwise, we risk calling alignment what is really just parallel improvisation.
Trust is not declared. It is built into systems.
Key Insight
South Asia’s pragmatic AI governance path will only become credible when soft-law coordination is backed by enforceable controls, sovereign infrastructure choices, and real labour protections.