Anthropic, an artificial intelligence (AI) developer, has significantly expanded the countries and institutions that can access the AI model ‘Mythos’, which has stood out in the security domain. South Korean government agencies and major companies such as Samsung and SK are reportedly included among the participants.

Anthropic stated on the 2nd (local time) that it has increased eligibility for participation in the cybersecurity consortium ‘Project Glasswing’, which is based on Mythos, to roughly 150 new institutions across more than 15 countries.

Project Glasswing is a security consortium formed when Anthropic disclosed in April that it had developed the Mythos model, and out of concern about potential misuse by hackers, Anthropic initially provided the Mythos model only to a little over 50 institutions, including financial firms.

Although Anthropic did not release a separate list this time, it explained that institutions in industry sectors not previously included, such as power, healthcare, and telecommunications, have been added.

Companies joining as partners reportedly include Samsung Electronics, SK hynix, and SK Telecom, which recently became Anthropic investors. Previously, Samsung Electronics and SK hynix participated as strategic infrastructure partners in Anthropic's Series H investment round. The scale of the two companies' investments is said to reach the trillion-won level.

The Korea Internet & Security Agency (KISA), an agency under the Ministry of Science and ICT, is also reportedly included among those eligible to participate. Earlier, the government decided to join GTAC, OpenAI's cybersecurity consortium.

With access to ‘Mythos’ now secured, cybersecurity capabilities are expected to be further strengthened. Anthropic said that, after reviewing the security of the initial group of a little over 50 partners, it found more than 10,000 security flaws with severity ratings of ‘High’ and ‘Critical’.

Anthropic said, “Within 6∼12 months, several other AI companies will also have models on par with Mythos, and they may release such models without safeguards,” adding, “Security personnel need to prepare response measures accordingly.”