The Indian Computer Emergency Response Team (CERT-In), India’s national cybersecurity agency under the Ministry of Electronics and Information Technology (MeitY), has gained access to Anthropic’s unreleased Mythos AI model, IT Ministry sources said on 3 November 2026. The agency will use the model to find vulnerabilities in critical systems and recommend patches, starting slowly and cautiously.

The access closes a months-long effort that began when India engaged the US and Anthropic over Mythos access in April 2026.

What CERT-In will do with Mythos

Mythos is a frontier model with advanced capabilities in uncovering software security flaws. Anthropic has kept it unreleased and shares it only with a small group of trusted organisations through Project Glasswing, its initiative to find and fix weaknesses in critical IT systems before attackers can exploit them.

According to government sources quoted by PTI, CERT-In will begin by experimenting with the model to understand its capabilities, applications and risks before any wider use. “We will slowly route anything we want to test. So it will be a slow process,” the sources said, adding that “the cure should not be worse than the problem.”

• Identify vulnerabilities in systems that may be exposed to attack
• Recommend patches and fixes for the weaknesses it finds
• Strengthen India’s response to emerging AI-driven cyber threats

Why cloud access is a concern

Mythos is available only as a cloud service, and officials are wary of sending sensitive details about critical systems to an external platform. To limit that exposure, CERT-In is also running several other AI models for vulnerability detection, which sources said deliver around 80 to 85% of Mythos’s capabilities.

The caution mirrors CERT-In’s own recent warnings. Earlier this week, the agency flagged AI agents reading Indian firms’ data undetected, underlining how the same AI capabilities that defend networks can also be turned against them.

How it fits India’s sector-led AI rules

The same sources backed sector regulators taking the lead on AI rules for critical infrastructure, pointing to the TRAI draft guidelines for responsible AI in telecom networks. Rules can be tighter in high-risk sectors, they said, while a single tight MeitY framework would affect everyone.

MeitY is separately preparing a consultation paper on AI regulation covering AI safety, deepfakes and a human-first approach, following Minister Ashwini Vaishnaw’s remark that the draft AI regulation paper is likely by November.

For Anthropic, India is a growing market: the company, which reported $17 billion in Q3 revenue, has named India as Claude’s second-largest market.

Key takeaways

• CERT-In now has access to Anthropic’s unreleased Mythos model for cyber defence.
• Use will be slow and calibrated, starting with tests of capability and risk.
• Cloud-only access raises data concerns, so CERT-In also uses other AI models.

FAQ

What is Anthropic’s Mythos model?

Mythos is an unreleased Anthropic AI model with advanced abilities to find cybersecurity vulnerabilities, shared only with trusted organisations under Project Glasswing.

How will CERT-In use Mythos?

CERT-In will use it to identify weaknesses in vulnerable systems and suggest patches, routing test cases to the model gradually.

Featured image: Anthropic CEO Dario Amodei at TechCrunch Disrupt 2023. Photo: TechCrunch via Wikimedia Commons (CC BY 2.0)

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter.

"Exciting news! Elets technomedia is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest insights!" Click here!