South Korea making its own security-centric AI model
Adapting existing local LLM project for security and sovereignty purposes and hopes to one day match Mythos
By The Register
South Korea is developing its own security-focused artificial intelligence model by adapting an existing domestic large language model project for national security and sovereignty purposes, with officials aiming to eventually match the capabilities of the Mythos system. Vice Prime Minister and Minister of Science and ICT Bae Kyung-hoon announced the plan on July 16 during a presidential policy briefing at the Blue House, stating the government will release a security-specialised proprietary AI model within the year .
He explained that the new model will be created by further training the country's existing proprietary AI model with security-related data to address cyber threats that exploit AI to find and use security vulnerabilities . Minister Bae said at the briefing held at Cheong Wa Dae, presided over by President Lee Jae Myung, that AI can be used to easily find and exploit security weaknesses, so preparation is needed on the defence side .
The initiative represents a push to make South Korea a world number two in AI security and to deploy a specialised sovereign AI model dedicated to cybersecurity by the end of 2026 . This move follows earlier regulatory steps, including the July 7 publication of an AI Security Red Teaming Guide and an AI Security Threat Response Manual by the Ministry of Science and ICT and the Korea Internet & Security Agency .
The government also plans to release a free general-purpose chatbot service based on a domestic AI model within the year, minimising usage limits and including features to help users find youth subsidies or welfare benefits .