South Korea Secures Access to Anthropic's Mythos Model
Reuters reports that South Korea's Science Ministry said the Korea Internet and Security Agency (KISA) secured access to Anthropic's cybersecurity model Mythos through Project Glasswing, the initiative aimed at using frontier AI to find and fix software vulnerabilities. The move follows Anthropic's June 2 expansion of Glasswing to about 150 organizations in more than 15 countries. Per the Financial Times and Anthropic, Korean participants also include Samsung Electronics, SK Hynix, and SK Telecom; Reuters notes Samsung declined to comment and SK did not immediately respond. The Korea Times reports KISA will use the model to strengthen national cyberdefense, and the ministry said South Korea will keep building domestic AI-based information-security capabilities. Anthropic has said partners across the program have already found more than 10,000 high- or critical-severity flaws.
What happened
Reuters reports that South Korea's Science Ministry said the Korea Internet and Security Agency (KISA) secured access to Anthropic's cybersecurity model Mythos by joining Project Glasswing, Anthropic's program for using frontier AI to identify and help fix software vulnerabilities. The ministry said it had been working with Anthropic and confirmed KISA's participation. The access aligns with Anthropic's June 2 announcement that it is extending Glasswing to roughly 150 new organizations in more than 15 countries.
Korean participants
Per the Financial Times and Anthropic, other Korean participants include Samsung Electronics, SK Hynix, and SK Telecom. Reuters reports Samsung declined to comment and SK did not immediately respond. CNBC and the Korea Times list the wider Glasswing group as including Okta, NATO, and the EU cybersecurity agency ENISA alongside major vendors.
Why it matters
A national cyber agency gaining hands-on access to a frontier security model reflects deepening public-private collaboration on operational cyberdefense. Anthropic has said partners across the program have already surfaced more than 10,000 high- or critical-severity flaws, signaling how quickly such models can find vulnerabilities at scale.
What to watch
- •Any published evaluation results or red-team findings on Mythos performance in government use.
- •The scope and governance of data sharing between KISA and private partners.
- •Ministry guidance on domestic AI-based information-security technologies.
Key Points
- 1KISA gained access to Anthropic's Mythos via Project Glasswing, giving a national cyber agency formal use of a frontier vulnerability-finding model.
- 2The access is part of Anthropic's broader expansion to about 150 organizations in 15-plus countries; Korean participants also include Samsung, SK Hynix, and SK Telecom, per the FT and Anthropic.
- 3For practitioners, government adoption of model-driven security tooling raises practical questions of validation, false-positive rates, data governance, and incident-response integration.
Scoring Rationale
A national cybersecurity agency securing access to a frontier vulnerability-finding model is a notable operational and geopolitical development for security practitioners, though it is a country-level facet of Anthropic's broader Glasswing expansion rather than a model-release-level event. That supports a mid-6s score.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
