TOKYO, June 18 (Reuters) – Japanese lenders may be forced to suspend services such as ATMs and online banking if sophisticated AI models end up posing a serious threat to the banking system, said the chair of the country’s banking lobby.
Frontier AI systems, such as Anthropic’s Mythos, can quickly identify vulnerabilities in software systems, sparking much concern about the potential for a barrage of cyberattacks.
“There are concerns about an increase in sophisticated cyberattacks that go beyond what has been anticipated,” Masahiko Kato, chair of the Japan Bankers Association and president of Mizuho Bank, told a press briefing.
“Certain services such as ATMs could be proactively suspended in order to protect customers’ assets,” he said.
Anthropic warned at Mythos’ launch in April that the product had uncovered thousands of software vulnerabilities — including flaws across every major operating system and browser — and said the fallout from its spread could be severe.
Banks have tightened checks on such AI tools and the U.S. government last week ordered Anthropic to suspend access to its frontier AI models for foreign nationals on national security grounds.
But inside the cybersecurity world, the reaction has been more measured — with some saying the broader response has been overblown, and that access to a Mythos-level large language model will not immediately enable hacking operations previously out of reach for bad actors.
(Reporting by Miho Uranaka; Writing by Anton Bridge; Editing by Edwina Gibbs)
