AI4Bharat is a non-profit, open-source institution and research lab incubated at the Indian Institute of Technology (IIT) Madras. It was founded in 2019 to address the problem of India's vast linguistic diversity and the lack of AI models capable of effectively processing and understanding Indian languages. The core objective is to bring parity in AI technologies for Indian languages with English.
The mechanism of AI4Bharat centers on creating and curating large-scale, public, open-source datasets and foundational AI models across all 22 scheduled Indian languages. Key outputs include the language model IndicBERT, which supports 12 major Indian languages, and the IndicTrans2 models, released in May 2023, which facilitate translation between all 22 scheduled languages and English. It also develops practical applications, such as the Supreme Court Vidhik Anuvaad Software (SUVAS) for translating judicial judgments.
AI4Bharat is closely connected to the government's digital initiatives, serving as the official Data Management Unit (DMU) for the Digital India Bhashini project under the Ministry of Electronics and Information Technology (MeitY). The Nilekani Center at AI4Bharat, launched on July 28, 2022, at IIT Madras, provides significant support for this work. Recently, in August 2026, AI4Bharat launched the Voice of India platform with Josh Talks AI to provide India-specific evaluation benchmarks for AI systems, moving beyond Western-centric models.