Explainer · Sovereign cyber AI
Project Glasswing is locked. India is mostly outside it.
Anthropic's most capable cyber model, Mythos, is available only through Project Glasswing — a vetted programme of about 200 organisations. No Indian organisation is a named partner. Here is what that means, and India's answer.
An independent explainer. We are not affiliated with Anthropic or Glasswing.
The facts
What Glasswing is, from the record
Sourced from Anthropic's own pages and named reporting.
What Mythos is
Claude Mythos is Anthropic's frontier, cyber-capable model, announced on 7 April 2026. Anthropic describes it as a general-purpose model whose offensive-security ability emerged from broader gains in coding and reasoning — not a narrow security tool.
It is deliberately restricted
Anthropic withheld Mythos Preview from public release, judging it too capable at finding and exploiting high-severity vulnerabilities. Access is only through Project Glasswing, its vetted-partner programme.
Who gets in
Glasswing launched with a dozen named partners (large cloud, security and finance firms) and about 40 more, and expanded by roughly 150 organisations across 15+ countries in June 2026 — around 200 vetted organisations in total. Access cannot be bought.
Where India stands
No Indian organisation is a named Glasswing partner. India is reported to be among the countries in the June expansion, but wider access for Indian institutions remains uncertain. India is, in effect, largely on the outside of the most capable cyber AI.
Where India stands
Outside the programme — and responding
The government reacted
As reported by Inc42 and the Economic Times (citing officials), the Centre has asked Sarvam AI and BharatGen to build “Mythos-like” cyber-AI models, intended to run on the government's isolated compute. This is reporting, not a published government mandate.
CERT-In is testing substitutes
India's IT Secretary S. Krishnan said (July 2026, as reported) that CERT-In runs other models while it awaits Mythos access, estimated at around 60–70% of Mythos's capability. That figure describes India's own interim government toolchain — not any Codesnag or VirtualVakil product.
The sovereign push is official
The direction of travel is clear in the public record: critical cyber work needs AI that India controls, hosted in India, under Indian law.
A Mythos alternative for India
The sovereign option available to Indian organisations.
Indian AI models vs Mythos
Sarvam, BharatGen, Krutrim — and the empty lane.
Sovereign AI for India
Why sovereignty, and what we are building.
Response Centre
Free public check-and-fix pages for critical vulnerabilities.
Questions
Straight answers
What is Project Glasswing?
Anthropic's vetted-partner programme, launched 7 April 2026, for access to its restricted Mythos cyber model. Access cannot be bought; participants must meet Anthropic's security requirements and the work is defensive.
How many organisations are in it?
Around 200 — a dozen named launch partners plus about 40 more at launch, then a further ~150 organisations across 15+ countries in a June 2026 expansion.
Is any Indian organisation a Glasswing partner?
No Indian organisation is a named partner. India is reported to be among the countries in the June expansion, but wider access for Indian institutions remains uncertain.
Is Codesnag part of Glasswing?
No. We are not affiliated with Anthropic or Glasswing. We build a separate, India-hosted sovereign alternative for cyber and legal.
Mythos is a product of Anthropic. Codesnag and VirtualVakil are not affiliated with, endorsed by or powered by Anthropic or Mythos, and we make no claim of benchmark parity with it. The name is used here only to describe the national conversation.