India is negotiating access to advanced AI models for cybersecurity and code security.
Key Highlights
- The government has prioritised access to Anthropic’s Mythos AI model.
- Talks are underway with US authorities and AI companies.
- CERT-In is using alternative AI models with 60–70% of Mythos’ capabilities.
- The models are being used to detect software vulnerabilities.
- Sensitive government code will require on-premises AI deployment in India.
New Delhi: The Indian government has made access to Anthropic’s advanced Mythos AI model a priority as it strengthens cybersecurity capabilities and software vulnerability detection.
IT Secretary S. Krishnan said discussions are underway with US authorities and AI companies to obtain access to the frontier AI model, which has demonstrated advanced capabilities in identifying hidden software vulnerabilities.
Until access becomes available, CERT-In has established a secure sandbox environment using alternative AI models that offer around 60–70% of Mythos’ capabilities. These models are being used to identify and fix software vulnerabilities while preparing secure workflows for future deployment.
Krishnan said the government also wants highly sensitive source code to be processed through on-premises AI systems within India rather than cloud-based infrastructure to ensure data security and sovereignty.
