INTERVIEW: Japan’s Mythos Response “Must Involve Big Tech”

Tokyo, May 19 (Jiji Press)–Liberal Democratic Party cybersecurity affairs leader Masaaki Taira has stressed the importance of involving U.S. technology companies in Japanese government responses to emerging cyber threats posed by artificial intelligence model Claude Mythos. The ruling LDP’s Headquarters for National Cybersecurity Strategy, headed by Taira, called on the government in an emergency proposal to bolster cybersecurity measures, as Mythos, developed by U.S. startup Anthropic, is highly adept at identifying system vulnerabilities and may be used in cyberattacks. “We must involve Big Tech,” Taira said in a recent interview, referring to major U.S. technology companies. “When I met a senior Anthropic official Friday, I was told the company will cooperate in any way it can.” “However, (the risks) are spreading to areas beyond the scope of what companies can judge, such as diplomacy and security,” he added. “Collaboration between the prime minister’s office and the White House is necessary.” “Last year, when I was minister in charge of cybersecurity, I told the Diet (Japan’s parliament) that the world will see AI battle AI, but this reality came earlier than I had expected,” Taira said regarding the emergence of Mythos. He said the LDP was prompted to draw up an emergency proposal by news reports that the U.S. government discussed measures against cyber threats with Anthropic and financial institutions in mid-April this year. “We’ll first work on the financial industry, which is especially prone to being targeted,” he said. “The National Cybersecurity Office and the Japan AI Safety Institute will thoroughly support the Financial Services Agency.” “We also need to look at critical infrastructure as a whole,” he continued. “We can utilize the framework of a public-private council for active cyberdefense, led by the NCO and involving the AISI, the council and government agencies.” Taira expressed confidence that Tokyo can gain the cooperation of AI companies, saying, “The LDP established a team on AI three and a half years ago and has built relationships of trust with major companies such as OpenAI, Anthropic and Google.” “We distinguished ourselves from the European Union and its extensive regulations under the policy of becoming the world’s most AI-friendly country,” he went on. Taira observed that even some earlier AI models have similar capabilities to identify vulnerabilities, making it possible to take action without having access to Mythos. Anthropic allows only a limited number of companies and groups to access the AI model given the risks it poses. “It’s important not to be overtaken by AI developed in countries that do not share Japan’s values,” he stressed. “We are entering an age in which AI developers stop making models publicly accessible and instead provide access only to trusted partners to strengthen defense because of the risks posed by the models,” Taira said. “What’s important is to not be left out of this and to collaborate with like-minded countries in order to enhance our negotiating power.” END [Copyright The Jiji Press, Ltd.]