Japan's largest lenders are preparing for a wave of AI-driven cyberattacks expected this autumn, as the arrival of a powerful new hacking tool, Anthropic's new artificial intelligence model, Mythos, reshapes their digital defense plans.
Anthropic Mythos Sparks Bank Alarm
The country's top three banking groups, Mitsubishi UFJ, Sumitomo Mitsui and Mizuho, are bracing for cyberattacks tied to Anthropic's Mythos, according to a report published by Nikkei Asia on Wednesday.
The lenders are reworking digital strategies that until now had focused on cutting costs and lifting revenue.
Mythos is built to detect software vulnerabilities at speeds beyond earlier AI systems. Anthropic has restricted access to the model out of concern that misuse could destabilize critical financial infrastructure.
The Japanese megabanks are expected to receive defensive access through Anthropic's Project Glasswing program in the coming weeks.
Also Read: Security Experts Pour Cold Water On Claude Mythos Hacking Apocalypse
FSA Working Group Takes Shape
Japan's Financial Services Agency launched a public-private working group this month to coordinate the response, alongside the Bank of Japan, the National Cybersecurity Office and Japan Exchange Group.
Finance Minister Satsuki Katayama said the forum, made up of 36 entities including the Japanese units of Anthropic and OpenAI, will share threat assessments and draft contingency plans.
The group is chaired by Mizuho's Chief Information Security Officer Osamu Terai. Members will also examine procedures for handling newly discovered vulnerabilities.
Regulators see the autumn window as critical, since attackers may begin probing legacy bank systems once Mythos-class tools spread more widely.
Why Banks Cannot Wait
Analysts say the risk lies less in the model itself and more in the gap between AI capability and patching speed at large institutions. Bank legacy systems were never built for adversaries that can scan code at machine speed.
Anthropic has told regulators that the Mythos preview has already surfaced thousands of high-severity vulnerabilities in widely used software. The findings have rattled supervisors from Tokyo to Frankfurt.
Japan has been a steady target of ransomware crews for years, with attacks on the JAXA space agency exposing weak spots across the country's public networks. The new working group is the first time Tokyo has placed AI-enabled cyber risk on par with traditional financial-stability threats.
Read Next: Twenty One Capital Becomes Tether's Bitcoin Arm As SoftBank Walks Away