Anthropic Extends Claude Mythos Cyber Shield to India, 14 Nations
Anthropic has handed its most powerful model to 150 critical-infrastructure bodies across 15-plus countries, India among them. The catch for Indian security chiefs is sharper than it looks.
The News
Anthropic has widened access to Claude Mythos, the model it calls its most powerful, handing the system to 150 fresh organisations across more than 15 countries. The move, announced on 2 June 2026, also scales up Project Glasswing, a joint industry effort that turns the model loose on partner codebases to find and fix serious software flaws before attackers reach them.
India sits squarely on the participant list, named alongside Australia, Canada, France, Germany, Italy, Switzerland, the Netherlands, Spain, Belgium, Sweden, Japan, New Zealand and South Korea. The cohort leans heavily towards bodies that keep essential services running: power, water, healthcare, communications and hardware. Named partners include Okta, Samsung, SK Hynix, SK Telecom, the NATO alliance and ENISA, the European Union's cybersecurity agency.
The rationale is blunt. "What each partner has in common is that a successful attack on their codebase could be catastrophic," Anthropic said. By the company's own estimate, a breach at a single one of these organisations could ripple out to more than 100 million people.
It is a steep jump from early April, when just 50 partners were given access. Claude Mythos is built to surface thousands of zero-day vulnerabilities over a matter of weeks, work that would otherwise tie up human research teams for far longer. The expansion lands days after Anthropic closed a 65 billion dollar funding round at a valuation of nearly 1 trillion dollars, ahead of a planned market debut.
Why It Matters
This is the clearest sign yet that frontier models have crossed from writing code to defending it at national scale. For most of the past two years the security conversation around generative AI ran the other way, with worry about models helping attackers write malware faster. Pointing the same capability at vulnerability discovery flips that script, and doing it across power grids and hospitals turns a research demo into infrastructure policy.
The competitive subtext is hard to miss. OpenAI has already shipped a security-focused model of its own, and rivals are racing to be the trusted layer beneath sensitive systems rather than just another chatbot.
There is also a giveaway economics story here. A firm valued at close to a trillion dollars can seed frontier access among 150 chosen bodies long before any invoice arrives. The last time a vendor moved this aggressively to become default plumbing for critical systems, it reshaped markets for a decade.
Indian Angle
India's inclusion is not a footnote. The country runs one of the world's most attacked digital estates, and the sectors Anthropic is targeting map almost exactly onto the assets overseen by the National Critical Information Infrastructure Protection Centre and flagged repeatedly by CERT-In. The 2022 ransomware shutdown at AIIMS Delhi and recurring probes at state power utilities are exactly the catastrophic-codebase scenarios the programme is pitched against.
For Indian banks and insurers, the regulatory read is immediate. The RBI's cyber-resilience expectations and SEBI's tightening disclosure rules already push boards to prove they are hunting vulnerabilities proactively. An AI system that surfaces zero-days at speed is exactly the control auditors will soon start asking about.
There is a home-grown contrast too. Indian model builders such as Sarvam and Krutrim have focused on language and consumer use cases rather than security tooling, leaving a gap that foreign frontier labs are now filling inside Indian infrastructure. For the country's vast IT-services industry, which manages codebases worldwide, the same tooling could become a billable service, or a reason clients ask why it is not already standard.
FAQ
Is India part of this rollout?
Yes. India is named in the body of Anthropic's announcement alongside 14 other countries. The programme targets power, water, healthcare, communications and hardware, the same sectors that sit under India's critical-infrastructure protection regime.
What exactly is Claude Mythos?
It is Anthropic's most capable model, here pointed at partner codebases to find security flaws. The company says it can identify thousands of previously unknown zero-day vulnerabilities over several weeks.
How big is Anthropic financially?
The expansion follows a 65 billion dollar funding round at a valuation of nearly 1 trillion dollars, ahead of a planned listing. That scale lets the firm seed frontier access among 150 organisations.
Where can I read the original announcement?
TechCrunch reported the expansion on 2 June 2026. The full coverage is linked in the attribution note below.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.
Sources & Citations
Frequently Asked Questions
Is India part of this rollout?
Yes. India is named in the body of Anthropic's announcement alongside 14 other countries including Australia, Canada, France, Germany, Japan and South Korea. The programme targets organisations running power, water, healthcare, communications and hardware, the same sectors that sit under India's national critical-infrastructure protection regime.
What exactly is Claude Mythos?
Claude Mythos is Anthropic's most capable model. In this programme it is pointed at partner codebases to surface security flaws, and the company says it can identify thousands of previously unknown zero-day vulnerabilities over a span of several weeks, compressing work that human teams would take far longer to complete.
How big is Anthropic financially right now?
The expansion follows a 65 billion dollar funding round that placed Anthropic at a valuation of nearly 1 trillion dollars, ahead of a planned listing. That scale is what lets the firm give away frontier-model access to 150 organisations rather than charging premium enterprise rates for it.
Where can I read the original announcement?
TechCrunch reported the expansion on 2 June 2026. The full coverage, including the complete list of partner countries and named organisations, is linked in the attribution note at the foot of this article.