The government is actively working to access Anthropic’s advanced AI model, Mythos, to enhance cybersecurity and detect software vulnerabilities. IT Secretary S. Krishnan mentioned that discussions are ongoing with US authorities and the company to secure access to Mythos and other cutting-edge AI models. While negotiations are in progress, the government is already utilizing alternative AI models to identify and address software vulnerabilities promptly.
Mythos has proven effective in uncovering vulnerabilities in widely used software that were previously overlooked due to various reasons. Krishnan highlighted the importance of accessing Mythos and similar advanced AI systems to systematically identify and rectify weaknesses. The government has engaged with US counterparts and companies developing these AI models, although formal access procedures are still under negotiation.
In the interim, the Indian Computer Emergency Response Team (CERT-In) has set up a dedicated sandbox to test software code using alternative AI models. These substitute models offer substantial capabilities similar to Mythos, aiding officials in thorough testing before accessing the advanced system. CERT-In is actively identifying and addressing vulnerabilities in software code through this controlled environment, ensuring preparedness for Mythos integration.
Krishnan emphasized that despite facing challenges due to US export restrictions on advanced AI technologies, the government is making progress in addressing software vulnerabilities. The ongoing “dry run” exercise is preparing systems and processes for seamless integration once access to Mythos is granted.
