Japan Steps Up Defense Against Advanced AI, Including Mythos

By Megumi Fujikawa

TOKYO--Japan is stepping up efforts to defend against new threats posed by artificial intelligence, as officials warned about the risks from new models like Anthropic's Mythos.

Prime Minister Sanae Takaichi on Tuesday directed her digital minister and other relevant officials to devise measures to protect critical infrastructure from frontier AI models and identify cybersecurity vulnerabilities.

That adds to mounting concerns among global governments and monetary authorities about the danger AI presents to financial systems and public services that won't be able to defend against increasingly faster, more sophisticated cyber attacks.

New AI models like Mythos are capable of finding software vulnerabilities on their own, discovering and exploiting holes in security that developers aren't aware of, at a speed that humans cannot match.

Japan's finance minister, Satsuki Katayama, said she discussed the AI cybersecurity issue with Treasury Secretary Scott Bessent at their meeting in Tokyo on Tuesday.

"We shared the view that we must act in sync with our Western allies to prevent such tools from being weaponized by those who do not share our values," she said.

Hisashi Matsumoto, the minister in charge of Japan's cybersecurity, said Japan's government must figure out how to fend off risks from advanced AI models based on the assumption that the technology isn't yet available to the businesses and service providers that need to protect themselves.

"We intend to disclose these measures in the near future and hope to make them public as soon as possible," Matsumoto said.

The government has set up a working group to discuss the topic with the central bank, as well as the private sector, including major lenders and technology companies.

Analysis by the International Monetary Fund suggests that extreme losses caused by a cyber incident could trigger funding strains, raise solvency concerns and disrupt broader markets.

Heightened cybersecurity concerns have tempered the U.S.'s prior emphasis on rapidly deploying AI in the technological arms race against China. The White House is considering an executive order to implement a rigorous AI oversight process, prioritizing the safety and stability of high-capability models over unfettered speed.

Katayama said Bessent told her progress has been made in establishing protocols between AI developers and the Treasury.

"I believe it is very positive that these developments have been tracked at home," she said.

Write to Megumi Fujikawa at megumi.fujikawa@wsj.com

(END) Dow Jones Newswires

May 12, 2026 00:25 ET (04:25 GMT)