12 March 2026·PIB

Digital India BHASHINI Division: National Hub for Language Technologies Powers End-to-End AI at Population Scale

In this note

1. At a Glance

BHASHINI (BHASHa INterface for India) — National Language Translation Mission using AI/NLP to enable digital services in 22 scheduled Indian languages [2][4].
Implemented by Digital India BHASHINI Division (DIBD) under Digital India Corporation (DIC) — a Section 8 company of MeitY [2][4].
Now powers one of the world's largest public-sector AI systems by volume, running on India's first vendor- and cloud-agnostic AI Sovereign Cloud [1][3].
UPSC relevance: GS-II (governance, digital inclusion), GS-III (AI, indigenous tech, data sovereignty).

12 March 2026 PIB release: DIBD enabled open-sourced Sarvam models on BHASHINI, expanding population-scale AI capabilities [1].
Platform now manages an ecosystem of 350+ deeply optimised AI models on a sovereign MLOps pipeline [3].
Yotta Data Services completed deployment of BHASHINI's end-to-end sovereign AI cloud on Yotta's Government Community Cloud (GCC) and Shakti Cloud — entirely Indian GPU infrastructure [3].
Showcased at India AI Impact Summit 2026; VoicERA launched on BHASHINI national infrastructure [5].

Conceptualised under the National Language Translation Mission (NLTM), announced in Union Budget 2021-22 [2].
July 2022: MeitY held brainstorming with researchers/startups to shape BHASHINI strategy [4].
Launched by PM Modi in July 2022 as India's AI-led language translation platform [2].
Hosted under Digital India Corporation (DIC), MeitY, as DIBD [2].
Evolution: from translation APIs → 300+ models (2024) → 350+ models + sovereign cloud + Sarvam integration (2026) [2][3].

Full form: BHASHa INterface for India [2].
Parent ministry: Ministry of Electronics & Information Technology (MeitY) [1].
Implementing body: Digital India BHASHINI Division (DIBD), under Digital India Corporation (DIC) — a Section 8 company of MeitY [2].
Languages covered: English + 22 scheduled Indian languages (Eighth Schedule) [2].
Models hosted: 350+ optimised AI models, including open-sourced Sarvam family [1][3].
Flagship tools: Anuvaad (text), Chitraanuvaad (video), Lekhaanuvaad (document), Shrutlekh (speech-to-text), WTMC (website plugin) [2][4].
Infrastructure partner: Yotta — GCC + Shakti Cloud (Indian GPU stack) [3].

Scientific / Technological

First Indian vendor- and cloud-agnostic AI Sovereign Cloud; specialised MLOps pipeline on open-source stack [1].
Integrates Sarvam open-sourced LLMs — indigenous foundation models for Indic languages [1].
Operates 350+ optimised models across speech, text, OCR, translation [3].

Social

Targets digital inclusion — bridging linguistic divide for non-English speakers [2].
Deployed at Maha Kumbh 2025 for multilingual pilgrim services [4].
Shrutlekh offers real-time multilingual speech transcription for accessibility [4].

Strategic / Governance

Data sovereignty: language datasets, models, citizen interactions remain within Indian jurisdiction [3].
Aligned with IndiaAI Mission goals announced at India AI Impact Summit 2026 [5].

Economic

Creates ecosystem for Central/State agencies + startups to monetise Indic-language AI [2].
Open APIs lower cost of building vernacular digital products.

Administrative

Cross-cuts Centre–State; provides language stack as a public good (akin to UPI/Aadhaar for language).

Jan 2025: BHASHINI deployed at Maha Kumbh, Prayagraj for multilingual support [4].
2025: Shrutlekh launched for real-time speech-to-text across Indian languages [4].
Feb 2026: India AI Impact Summit; VoicERA launched on BHASHINI [5].
Feb 2026: Yotta–BHASHINI sovereign AI cloud deployment announced [3].
12 Mar 2026: Open-sourced Sarvam models integrated on BHASHINI [1].

BHASHINI = BHASHa INterface for India [2].
Implementing ministry: MeitY (not Ministry of Culture / Education) [1].
Run by Digital India BHASHINI Division under Digital India Corporation — a Section 8 company [2].
Launched under National Language Translation Mission, Union Budget 2021-22 [2].
Covers 22 scheduled Indian languages [2].
Hosts 350+ AI models (2026) [3].
Operates India's first vendor- and cloud-agnostic AI Sovereign Cloud [1].
Sarvam AI models integrated — open-sourced — March 2026 [1].
Infrastructure partner: Yotta Data Services (GCC + Shakti Cloud) [3].
Tools: Anuvaad (text), Chitraanuvaad (video), Lekhaanuvaad (document), Shrutlekh (speech) [2][4].
Deployed at Maha Kumbh 2025 for pilgrim services [4].
VoicERA launched on BHASHINI at India AI Impact Summit 2026 [5].

GS-II: e-governance, digital inclusion of linguistic minorities.
GS-III: indigenisation of technology, AI policy, data sovereignty, Digital Public Infrastructure (DPI).
Probable stems:
"BHASHINI exemplifies India's emerging template of Digital Public Infrastructure for AI. Examine." (GS-III)
"Discuss how language-tech sovereignty advances inclusive e-governance in India." (GS-II)
"Critically assess the role of open-source foundation models like Sarvam in India's AI strategy." (GS-III)

IndiaAI Mission — overarching AI policy framework hosting BHASHINI.
Digital Public Infrastructure (DPI) — BHASHINI as language layer of India Stack.
Eighth Schedule of the Constitution — basis of 22 scheduled languages.
National Language Translation Mission (NLTM) — parent programme.
Section 8 companies (Companies Act 2013) — governance form of DIC.
Data Protection Act, 2023 — sovereignty/locality interplay.
India AI Impact Summit 2026 — multilateral AI governance.
Sarvam AI / open-source LLMs — indigenous foundation-model ecosystem.

BHASHINI is under MeitY, not Ministry of Culture or Education.
DIBD sits under Digital India Corporation (Section 8 company) — not a statutory body or department.
BHASHINI ≠ NLTM; BHASHINI is the platform under the NLTM mission.
Sarvam models are integrated/open-sourced on BHASHINI — Sarvam itself is a private Indian AI firm, not a government entity.
Covers 22 scheduled languages, not "all" Indian languages.