US Tightens Security Measures on Frontier AI; South Korea Urged to Establish Risk Assessment Framework

The US has incorporated Frontier AI into national security management, prompting South Korea to develop a national risk assessment framework for Frontier AI and redesign its sovereign AI strategy. A report by the Institute for National Security Strategy (INSS) highlights how the MITOS incident—where Anthropic's model was halted due to security concerns—has elevated overseas access to Frontier AI as a critical national security variable. Trump's executive order on advanced AI innovation and security includes measures such as 30-day autonomous pre-screening, NSA-led classified benchmarking, and the establishment of an AI cybersecurity clearinghouse. The report notes this marks a symbolic shift of AI governance from industry policy to national security governance. Following the MITOS incident, the US demanded restrictions on Anthropic's models, leading to a partial service shutdown. The report argues that corporate self-regulation can influence regulatory effectiveness and warns that gaps in specific models may provide opportunities to competitors like China's Z.ai. South Korea is urged to develop a cyber-capability-focused risk assessment system, establish AI security governance under the National Intelligence Service (NIS), and reconfigure its sovereign AI strategy to include continuity of access.