The Indian government is actively negotiating access to Anthropic's frontier artificial intelligence model, Mythos, as part of a broader push to identify and fix software vulnerabilities across critical systems, IT Secretary S. Krishnan said on Monday, 13 July. Speaking in New Delhi, Krishnan confirmed that discussions are underway with US authorities and the company, even as India prepares parallel infrastructure using alternative AI tools.

Why Mythos Is a Priority

Krishnan highlighted Mythos's demonstrated capability to surface vulnerabilities in widely used software — weaknesses that had previously gone unaddressed because they were considered low priority, too costly to remediate, or had simply gone undetected. 'Mythos has demonstrated an ability to uncover vulnerabilities in widely used software that often remained unresolved because they were considered low priority, too expensive to fix, or had simply gone undetected,' he said.

The IT Secretary described this as a systemic opportunity: the ability to methodically scan codebases, flag latent risks, and correct them before they are exploited. Access to Mythos and comparable frontier AI systems has accordingly been designated a key government priority.

Negotiations Underway, But No Access Yet

'The government has taken up the matter with its counterparts in the United States as well as the companies developing these frontier AI models. However, access involves a formal process that is currently being negotiated,' Krishnan said. He acknowledged that prevailing US export restrictions on advanced AI technologies remain a significant hurdle in securing access to such systems.

This comes amid a broader global contest over frontier AI access, with several governments seeking preferential or early access to leading models for national security and infrastructure applications.

CERT-In's 'War Room' Already Operational

In anticipation of formal access, the Indian Computer Emergency Response Team (CERT-In) has established a dedicated sandbox — described internally as a 'war room' — where alternative AI models are being deployed to test software code, detect vulnerabilities, and develop secure workflows. Krishnan characterised the exercise as a 'dry run' designed to ensure that systems and processes are operationally ready once Mythos access is secured.

The substitute models currently in use are estimated to deliver approximately 60 to 70 per cent of Mythos's capabilities — sufficient, according to officials, to conduct extensive testing and address a meaningful share of software vulnerabilities even before the more advanced system becomes available.

Significance for India's Cybersecurity Posture

The initiative reflects a deliberate shift in India's cybersecurity strategy: moving from reactive patching to AI-assisted proactive vulnerability discovery. Notably, CERT-In is not merely waiting — it is actively using the sandbox environment to probe software code, identify weaknesses, and deploy fixes through a controlled, structured process.

The government's parallel-track approach — negotiating for Mythos while simultaneously building readiness with alternative models — signals that cybersecurity preparedness is being treated as a non-negotiable priority regardless of when formal access is granted. How quickly US export control negotiations conclude will likely determine the pace of the next phase.