South Korea wants its own Mythos-class security AI

South Korea is developing its own security-focused AI model and wants it online by the end of 2026 to give the country sovereign bug-finding capabilities.

Deputy Prime Minister and Minister of Science and ICT Bae Kyung-hoon disclosed the project and said it is needed so South Korea can field a bug-finding model to rival Anthropic’s Mythos. According to The Register, access to Mythos was blocked twice by the US government: first when Anthropic was required to offer it only to American citizens, a condition the company could not meet and so it blocked all access, and later when Washington ordered Anthropic to take the service down while it investigated allegations of potentially dangerous performance problems.

Those episodes, the publication said, pushed many countries to conclude that the US could eventually restrict access to powerful models, leaving US-based organizations and national security agencies with an advantage. Washington has since restored limited access to Mythos for some allies, but interest in sovereign AI has continued to rise.

South Korea’s plan is to adapt a locally developed frontier model by adding security-related information to its training corpus. Bae said the country now aims to build a Mythos-class system.

The government is also seeking bids for two additional projects:

Recommended reading

Linus Torvalds backs AI coding in Linux

• a chatbot that will be made freely available to all residents
• an agentic application designed to help people interact with government services

Bae made the comments during a policy briefing led by President Lee Jae Myung. The discussion also covered using AI to detect fake news in real time and to process complaints about government services faster than is currently possible.