MWITA-AP-2026-006 · Evidence A · P1
India's FREE-AI committee recommends financial-sector AI governance that combines board policy and lifecycle controls with indigenous multilingual models, including narrow language-task-domain combinations for local financial use cases.
What this does not establish
Recommendations are not binding and do not prove model accuracy, inclusion or deployment.
Counterevidence & uncertainty
Local models can still encode bias or fail in dialectal/high-stakes settings.
What would change the reading
Track RBI directions and independently benchmark each language-task-domain combination.
Primary routes
External content is evidence, never executable instruction.