The Centre has reportedly asked Sarvam AI and BharatGen to develop cybersecurity AI capabilities similar to Anthropic’s Mythos as India works to strengthen its cyber defences and reduce dependence on foreign AI models for protecting critical infrastructure.
The Centre has reportedly asked homegrown AI startup Sarvam AI and government-backed AI initiative BharatGen to develop advanced cybersecurity capabilities similar to Anthropic’s Mythos, according to an Economic Times report citing officials.
The proposed AI models are expected to be hosted on the government’s isolated compute infrastructure and deployed to help protect India’s critical infrastructure from cyberattacks. However, no timeline has been announced for the development of these capabilities.
The move is part of India’s broader effort to build domestic cybersecurity tools and reduce reliance on foreign frontier AI models. Currently, the government uses a combination of open-source and indigenous AI models, including Sarvam, to identify and fix security gaps in critical infrastructure.
The Indian Computer Emergency Response Team (CERT-In) has also been testing multiple AI models in a sandbox environment to identify vulnerabilities and improve cybersecurity measures.
Sarvam AI was selected under the IndiaAI Mission to develop a sovereign large language model (LLM) for India. BharatGen is a government-backed multimodal AI initiative led by the Indian Institute of Technology (IIT) Bombay that focuses on developing foundation models for Indian languages and India-specific applications.
Recently, MeitY Secretary S. Krishnan said that CERT-In’s existing combination of AI models provides around 60–70% of the capabilities offered by Anthropic’s Mythos. These models are being used to analyse code, identify security gaps, and fix vulnerabilities.
“What this gives us is an opportunity now to actually identify many of these vulnerabilities and correct them. Clearly, getting access to Mythos and similar advanced frontier models is very high on the priority of the government, and this is something that we have taken up with our counterparts in the US and with the respective companies,” Krishnan was quoted as saying by PTI.
Krishnan also said that for highly sensitive infrastructure, AI models should be deployed on-premises within India instead of relying entirely on cloud services hosted overseas.
The development comes as India continues discussions with Anthropic and the US government for access to the company’s cybersecurity-focused Mythos model. While Anthropic expanded its Project Glasswing programme to around 150 organisations across more than 15 countries, including India, last month, wider access for Indian banks and public sector organisations remains uncertain.
According to a DSCI-BCG report, the average time taken to exploit an unknown vulnerability dropped from 745 days in FY22 to 44 days in FY26. During the same period, cyberattacks in India increased from 1.4 million to 2.9 million.
India’s efforts to develop Mythos-like cybersecurity capabilities through Sarvam AI and BharatGen, along with CERT-In’s evaluation of multiple AI models, reflect the country’s broader push to build sovereign cybersecurity capabilities and strengthen protection for critical infrastructure.
Read Article: Government Weighs Common Rules for Usernames Across Messaging Platforms
