Korea joins Anthropic’s Project Glasswing for Mythos cybersecurity AI model

A smartphone displays the logo of the U.S. artificial intelligence safety and research company Anthropic, in Mulhouse, France, April 21. AFP-Yonhap

Korea is joining Anthropic’s cybersecurity initiative Project Glasswing, which will give the country access to the company’s advanced artificial intelligence (AI) model Claude Mythos as part of a broader expansion aimed at strengthening global cyber defenses.

Anthropic announced Tuesday (local time) that it is expanding the project, a collaborative cybersecurity program built around its Mythos model, to around 150 organizations across 15 countries, significantly widening access beyond its initial cohort of roughly 50 institutions. Mythos, unveiled in April, is a security-focused AI model designed to identify software vulnerabilities at an expert level.

Under the latest expansion, the Ministry of Science and ICT will participate in the program through the Korea Internet & Security Agency (KISA), the state-run agency for cyber incident response and digital security.

“We are aware that KISA has joined Project Glasswing and secured access to the Mythos model,” a ministry official said.

KISA is expected to be granted access to the Mythos Preview model and its related tools, such as Claude Security.

Industry sources also indicate that major Korean companies, such as Samsung Electronics, SK hynix and SK Telecom, are also expected to be included as new partners, though their participation has not been officially confirmed.

Both Samsung Electronics and SK hynix previously took part in Anthropic’s Series H funding round as strategic infrastructure partners. Their combined investment is said to be in the trillion‑won range, with Samsung’s commitment alone reportedly reaching several trillion won.

Since introducing Mythos, Anthropic has restricted access to vetted institutions due to concerns that such capabilities could be misused maliciously. Within Project Glasswing, participating organizations use the model for vulnerability detection, patching and pre-release security testing.

The company said early participants have already used Mythos to uncover more than 10,000 high-severity or critical vulnerabilities within weeks.

Anthropic framed the rapid expansion as a response to escalating AI‑driven cyber threats and a shrinking window for controlled deployment, arguing defenses must scale before similar models become widely available.

“Within 6 to 12 months, we expect that many other AI companies will have Mythos-class models, and they could release them without safeguards that prevent misuse,” the company said. “In that world, cyberattacks could occur much more often, and in much more unpredictable forms. It’s imperative that cyber defenders adapt to maintain pace.”