MUFG, SMBC, and Mizuho are projected to gain access to Anthropic’s Claude Mythos AI model as early as this month, according to Nikkei. This comes amid efforts by Anthropic to expand access to the model, which has demonstrated significant advancements in areas such as reasoning and cybersecurity vulnerability identification, despite facing pushback from the U.S. government. In Japan, financial authorities and leading banks are collaborating to form public-private working groups aimed at assessing risks and establishing standardized defenses against advanced AI systems like Claude Mythos.

MUFG: Mitsubishi UFJ Financial Group (MUFG) is one of Japan’s largest megabanks, providing comprehensive financial services including retail, corporate, and investment banking on a global scale. It has been actively integrating AI for applications like loan reviews and business co-creation models. In this development, MUFG is expected to secure access to Anthropic’s Claude Mythos AI model as early as this month to bolster cyber defenses through Japan-U.S. collaboration.
SMBC: Sumitomo Mitsui Banking Corporation (SMBC) operates as a major Japanese financial institution, recently launching the SMBC Connect brand for global transaction banking tailored to AI-driven treasury needs. It has partnered with Sakana AI to develop automated strategic proposal systems for wholesale banking advisory. SMBC is among the three megabanks gaining access to Anthropic’s Claude Mythos AI model starting this month.
Mizuho: Mizuho Financial Group is a leading Japanese bank offering diverse financial services, having recently selected Oracle Autonomous AI Database to enhance system protection. It has been involved in governmental discussions on AI cybersecurity risks posed by advanced models. Mizuho is set to obtain access to Anthropic’s Claude Mythos AI model imminently amid efforts to strengthen defenses.
Anthropic: Anthropic is a U.S.-based AI startup developing the Claude family of frontier models with a focus on safety, advanced reasoning, and agentic capabilities. In April 2026, it released Claude Mythos Preview, recognized for exceptional performance in cybersecurity vulnerability discovery and complex multi-step execution. Anthropic is expanding selective access to Claude Mythos, including to Japanese megabanks like MUFG, SMBC, and Mizuho, despite opposition from the White House.

`json
{
“Model Capabilities”: “Claude Mythos Preview shows advancements in reasoning, planning, and multi-agent coordination.”,
“International Access”: “Anthropic is preparing to expand Claude Mythos access to organizations in Japan.”,
“Japan Cybersecurity Response”: “Japanese financial authorities and top banks are forming working groups to verify risks and standardize defenses against advanced AI models.”
}
`