Japan's 3 megabanks set to use Anthropic's latest AI for cyber defense

TOKYO - Japan's three megabanks, including MUFG Bank, are expected to gain access to U.S. startup Anthropic's latest artificial intelligence model Claude Mythos, sources familiar with the matter said Wednesday.

The move comes after the Japanese government is believed to have sought access to the AI model, as Tokyo and Washington pledge to work together to respond quickly to cyberattacks targeting vulnerabilities in financial systems.

While Claude Mythos, released in April, is widely viewed as reshaping the AI industry long led by ChatGPT developer OpenAI, Anthropic's model has sparked fears it could be exploited for attacks on a wide range of systems.

Executives from the banks -- MUFG Bank, Sumitomo Mitsui Banking Corp., Mizuho Bank -- attended a meeting with U.S. Treasury Secretary Scott Bessent during his visit to Japan earlier this week, and access to Claude Mythos may have been among the topics discussed.

Amid safety and security concerns, access to the Claude Mythos model is currently restricted to IT companies such as Google LLC and some financial institutions, with Anthropic saying its preview "has already found thousands of high-severity vulnerabilities."

Japan's Financial Services Agency has decided to set up a framework for public- and private-sector officials to stronger responses to Claude Mythos, consulting with the Bank of Japan and major banks.

A working group is likely to meet as early as Thursday to information-gathering efforts and the direction of steps to enhance security, the sources said. Anthropic was established in 2021 by former OpenAI employees.