Anthropic Brings Its Most Powerful Mythos AI Model to India; Expands Project Glasswing To 14 Other Countries
Anthropic said it is expanding Project Glasswing, its restricted cybersecurity programme built around Claude Mythos Preview, to about 150 new organisations across 14 additional countries, more than tripling the initial early-April cohort of ~50. Anthropic’s blog and the Financial Times cite partners including Samsung, SK Telecom, SK Hynix, Okta, Swift, Euroclear and ICE, plus NATO and ENISA. The company says the model is kept non-public due to exploit-building capabilities.
Partner expansion could increase Okta’s security tooling effectiveness and highlight demand for AI-assisted vulnerability discovery.
Okta is named as a new Project Glasswing partner, gaining early access to Anthropic’s restricted Mythos Preview for cybersecurity testing.
Limited direct near-term impact; any reaction likely tied to broader AI/security sentiment rather than fundamentals.
Background
Project Glasswing is Anthropic’s restricted cybersecurity initiative using its most capable (publicly unavailable) Claude Mythos Preview model to help partners find and remediate vulnerabilities.
Why it matters
The expansion triples partner count and adds new countries, with the stated rationale that successful attacks could affect 100M+ people. This is primarily a cyber-risk mitigation story rather than a direct financial guidance event for partners.
Market relevance
Read-through is to the AI-security ecosystem: broader deployment of frontier-model cyber testing can support demand for security tooling, but the article provides no partner-specific financial impact.
Market effects
Reinforces a growing market for AI-assisted vulnerability discovery and managed security workflows; may increase competitive pressure on cybersecurity vendors and cloud security platforms.
India inclusion broadens the geographic footprint of advanced AI security testing, potentially increasing local enterprise security spend over time.
Partners span critical infrastructure and financial plumbing (telecom, payments/messaging, exchanges), underscoring systemic cyber risk mitigation using frontier models.
Alternative perspectives
Because the model remains restricted and the article lacks commercial terms, the partnership may be more PR/defensive than a revenue catalyst for named vendors.
Key missing details are contract size, cost, and whether findings translate into measurable product adoption or renewals; also, defensive cybersecurity announcements often have muted equity impact.
Key entities
- AI companyAnthropic
Expanding Project Glasswing to ~150 new organizations across multiple countries using Claude Mythos Preview.
- programProject Glasswing
Restricted cybersecurity initiative pairing partners with Anthropic’s advanced model for vulnerability discovery.
- AI modelClaude Mythos Preview
Most capable, restricted Anthropic model used for code analysis and exploit construction capabilities.
