The Indian government has asked AI unicorn Sarvam AI and BharatGen to develop advanced cybersecurity AI models similar to Anthropic's Mythos, aiming to reduce reliance on foreign frontier models for protecting critical infrastructure, according to inc42.com. These models will be hosted on the government’s isolated compute infrastructure and deployed to safeguard critical infrastructure from cyberattacks. No timeline has been specified for the development of these capabilities.
Currently, the government employs a mix of open-source and indigenous AI models, including those from Sarvam AI, to detect and address security vulnerabilities in critical infrastructure. The Indian Computer Emergency Response Team (CERT-In) has been testing alternative AI models in a sandbox environment to identify weaknesses. MeitY secretary S Krishnan stated that CERT-In’s existing combination of models delivers about 60-70% of Mythos’ capabilities and is actively used to probe code, find security gaps, and fix vulnerabilities.
The initiative reflects India’s effort to enhance its cybersecurity posture by developing domestic AI solutions that can match advanced frontier models like Mythos. Access to such models is a high priority for the government, which has engaged with US counterparts and relevant companies to gain access. This move aligns with broader efforts to secure critical infrastructure against increasingly sophisticated cyber threats using AI technologies.
CERT-In’s ongoing work with indigenous AI models and the government’s collaboration with Sarvam AI and BharatGen represent concrete steps toward strengthening national cybersecurity. MeitY secretary S Krishnan’s comments highlight the strategic importance of these efforts, which currently deliver a significant portion of Mythos-level capabilities in vulnerability detection and mitigation.