Anthropic Mythos: Glasswing Adds 150 Orgs in 15+ Countries

Abhishek GautamAbhishek Gautam10 min read
Anthropic Mythos: Glasswing Adds 150 Orgs in 15+ Countries

Quick summary

June 2, 2026: Anthropic expanded Project Glasswing — Claude Mythos Preview to ~150 new orgs across 15+ countries. Power, water, health, comms. 10,000+ critical flaws found since April.

Anthropic expanded Project Glasswing on June 2, 2026, giving ~150 additional organizations in more than 15 countries access to Claude Mythos Preview — the restricted model the company says found thousands of zero-days and helped early partners surface 10,000+ high- or critical-severity flaws since April.

Sectors added in this wave include power, water, healthcare, communications, and hardware — industries underrepresented in the first ~50-partner cohort that included AWS, Microsoft, Nvidia, Google, and US government teams.

What Is Project Glasswing?

Glasswing (named after the transparent-wing butterfly) is Anthropic's defensive industry program for scanning critical codebases with Mythos Preview — a non-public frontier model Anthropic says exceeds skilled humans at finding and exploiting vulnerabilities.

Anthropic committed up to $100M in Mythos usage credits plus $4M to open-source security orgs in the original launch.

Mythos will not ship to the general public — Anthropic cites misuse risk without adequate safeguards.

Public teams get Claude Security on Opus 4.8, which Anthropic said patched 2,100+ vulnerabilities in three weeks — a separate, safer product lane.

For the April Mythos launch and White House friction, see Anthropic Mythos White House Blocked Rollout.

Who Joined the June 2 Expansion?

Anthropic did not publish the full list. Reporting (Financial Times, TechCrunch) cited examples including:

  • Okta (identity)
  • Samsung, SK Hynix, SK Telecom (South Korea)
  • NATO
  • EU ENISA — see Anthropic Mythos EU ENISA access (June 2026 invite)
  • Countries span US-allied markets: Australia, Canada, France, Germany, Italy, Japan, India, South Korea, and others

CyberScoop reported Rubrik among vendors gaining access in this round.

Developer and Security Team Implications

Two-tier security products: Mythos-class access is vetted-partner only; day-to-day engineering uses Opus + Claude Security — know which tier your vendor contract actually includes.

Supply-chain blast radius: Many new partners maintain code other orgs embed — a flaw in their repo becomes your incident.

IPO timing: Expansion landed the same week as Anthropic's confidential IPO filing and Trump's voluntary frontier-model review order.

Same-day policy tension: Governments want early model access for cyber review while restricting public release of the strongest offensive-capable weights — plan for compliance bifurcation.

Key Takeaways

  • June 2, 2026: ~150 new Glasswing orgs, 15+ countries, critical infrastructure sectors
  • 10,000+ high/critical findings cited from ~50 early partners since April
  • Mythos stays non-public; Claude Security on Opus 4.8 is the broadly available path
  • Named examples include Okta, Samsung, SK Hynix, NATO, ENISA (per press)
  • For developers: demand clarity on Mythos vs Opus in contracts; treat upstream vendor code as attack surface

Sources

FAQ

Frequently Asked Questions

What is Anthropic Project Glasswing?

Project Glasswing is an Anthropic-led initiative that gives vetted organizations access to Claude Mythos Preview to find and fix critical software vulnerabilities in infrastructure and enterprise codebases. It launched in April 2026 with about 50 initial partners.

How many organizations joined Glasswing on June 2, 2026?

Anthropic announced approximately 150 additional organizations across more than 15 countries gained access to Claude Mythos Preview on June 2, 2026, expanding beyond the initial cohort of about 50 partners.

Will Claude Mythos be released to the public?

Anthropic has said it does not plan to release Mythos-class models to the general public due to misuse concerns without sufficient safeguards. It offers Claude Security using Claude Opus 4.8 for broader defensive use.

What vulnerabilities did Project Glasswing find?

Anthropic reported that early Glasswing partners collectively identified more than 10,000 high- or critical-severity security flaws using Claude Mythos Preview in the weeks after the April 2026 launch.

Free Weekly Briefing

The AI & Dev Briefing

One honest email a week — what actually matters in AI and software engineering. No noise, no sponsored content. Read by developers across 30+ countries.

No spam. Unsubscribe anytime.

More on Cybersecurity

All posts →

Written by

Software Engineer based in Delhi, India. Writes about AI models, semiconductor supply chains, and tech geopolitics — covering the intersection of infrastructure and global events. 795+ posts cited by ChatGPT, Perplexity, and Gemini. Read in 164 countries.