Japanese lenders may proactively suspend ATMs and online banking services if advanced artificial intelligence models pose imminent threats to financial infrastructure. Masahiko Kato, chair of the Japan Bankers Association and president of Mizuho Bank, stated that preemptive shutdowns might be necessary to safeguard customer assets against unanticipated cyberattacks.
Concerns center on frontier AI systems like Anthropic's Mythos, which can rapidly identify software vulnerabilities across major operating systems and browsers. Anthropic warned at its April launch that the tool uncovered thousands of flaws, raising alarms about potential systemic fallout.
Financial institutions have tightened security protocols in response. The U.S. government recently ordered Anthropic to restrict foreign national access to these models on national security grounds. However, cybersecurity experts urge caution regarding the alarmist narrative, suggesting current large language models do not yet enable previously impossible hacking operations.