A federated architecture for sector-led AI governance: lessons from India

Researchers Avinash Agarwal and Manisha J. Nene have published a paper proposing a comprehensive "whole-of-government" architecture to address potential fragmentation in India's sector-led AI governance strategy. While India's current vertical, sector-specific approach promotes innovation, the authors argue it risks creating policy silos and inconsistent implementation. Their solution applies a five-layer conceptual framework to create two actionable architectures: a primary model assigning clear governance roles to India's key institutions, and a detailed federated system specifically for national AI Incident Management.

The federated incident management architecture directly tackles the data silo problem by implementing a common national standard. This allows different sectors—such as healthcare, finance, and transportation—to maintain their own data collection practices while enabling cross-sectoral analysis for risk assessment and policy development. The system facilitates aggregation of incident data across borders without requiring centralized control, creating a globally relevant template for nations pursuing similar governance models. The paper demonstrates the architecture's practical utility through a detailed case study, showing how it connects high-level policy goals with operational implementation plans.