[BODY]
Masayoshi Son, chairman of SoftBank Group, warned at a corporate event held in Tokyo, Japan, on the 16th that artificial intelligence (AI)-based cyber threats represent "Japan's worst crisis since the arrival of the Black Ships." Just as the entry of American warships into Tokyo Bay in 1853 exposed the limits of the shogunate system, his remarks suggested that AI-powered cyberattacks could shake Japan's entire security, financial, and industrial systems.
Son's warning came immediately after the United States blocked access to Anthropic's cutting-edge security AI, "Mythos." The remarks reflected anxiety on the Japanese side, which had sought to use Mythos—capable of swiftly identifying software vulnerabilities and analyzing attack vectors—for cyber defense. In particular, the fact that even Japan, which has maintained close ties with the U.S. in both economic and security matters, became subject to export controls heightened the sense of crisis, reinforcing the perception that even allies are no exception when it comes to cybersecurity.
With it now confirmed that access to cutting-edge AI can be restricted at any time depending on diplomatic and security circumstances, the need to secure AI sovereignty and AI security capabilities is growing not only in Japan but also among U.S. allies such as South Korea and Europe. As AI emerges as a core tool for cyber defense—including vulnerability detection, threat analysis, and attack simulation—analysts say AI capability itself has begun to determine national competitiveness.
According to the information technology (IT) industry on the 22nd, recent cyberattacks, powered by AI, are evolving beyond simple hacking to target national infrastructure, financial networks, communications networks, and defense systems. This has put governments around the world on alert.
The United Kingdom is a representative example. According to the UK's National Cyber Security Centre (NCSC), cyber incidents related to critical infrastructure such as energy, healthcare, transportation, and finance exceeded 200 over the one-year period through May this year. About 75% of these were estimated to be linked to state actors such as Russia, China, and Iran. The NCSC warned that advances in AI technology could further increase the speed and scale of cyber threats.
For governments, using AI as a national security asset has now become a necessity rather than a choice. The U.S. Department of Defense is introducing generative AI into intelligence analysis, operational support, and software development, while the UK's Government Communications Headquarters (GCHQ) and NCSC are working to strengthen capabilities to protect critical infrastructure using AI. The U.S. Defense Advanced Research Projects Agency (DARPA) is also pursuing, as a national security task, the development of technology in which AI automatically detects and fixes software vulnerabilities.
However, as a structure forms in which the use of AI itself determines security capabilities, concerns are growing that an "AI security gap" between nations could materialize. Even with the same security personnel and budget, a significant difference inevitably arises in the cyber defense capabilities of countries that can use cutting-edge AI and those that cannot.
In particular, as the United States begins to regard advanced AI models as strategic assets and control access to them, some analysts say that access to AI itself is emerging as a new form of competitiveness. With AI models now being managed as strategic assets following semiconductor export controls, observers point out that if a situation in which only certain countries can use cutting-edge security AI becomes reality, the gap in cybersecurity capabilities between nations could widen.
South Korea is not free from this situation either. The government is expanding cooperation on AI safety and cybersecurity with global AI companies such as OpenAI and Anthropic, but the right to use cutting-edge AI can be restricted at any time according to U.S. security judgments. For South Korea, which holds critical infrastructure such as finance, public services, telecommunications, and defense, voices are growing that it needs to secure independent AI security capabilities separate from the use of foreign frontier models.
Domestically, the movement to view AI as a national security asset is also gaining momentum. Naver recently established a dedicated organization for defense AI transformation (AX) and has begun building a defense AI platform that integrates and analyzes military data. It is also developing technology that can be applied to drones, unmanned systems, and tactical vehicles using a defense-specialized AI model based on HyperCLOVA X. The government is also accelerating efforts to secure AI sovereignty by pursuing an independent AI foundation model project and the construction of a defense AI platform.
Experts assess that this Mythos access restriction incident demonstrates the importance of "AI security sovereignty" beyond mere technological control. They say that while past competition to secure semiconductors determined industrial competitiveness, access to cutting-edge AI and independent AI security capabilities will become key factors determining national cyber defense capabilities and security competitiveness going forward.
