TOKYO, May 12 (Reuters) - Japan will establish a

public-private working group this week to address cybersecurity

risks to the financial system posed by Anthropic's new

artificial intelligence model Mythos, Finance Minister Satsuki

Katayama said on Tuesday.

* Katayama, who met U.S. Treasury Secretary Scott Bessent in

Tokyo on Tuesday, said the group aims to develop shared

understanding of potential threats arising from AI among

financial and tech industries and policymakers, with inputs from

the U.S. government.

* A total of 36 entities, including financial institutions

from megabanks to internet banks, the Bank of Japan and the

Japanese units of Anthropic and OpenAI, will hold the group's

first meeting on Thursday, chaired by Mizuho Financial Group's ( MFG )

Chief Information Security Officer Osamu Terai,

according to the Financial Services Agency (FSA).

* Anthropic has launched Project Glasswing to offer access

to its latest AI model for defensive purposes to a limited

number of organisations. The FSA said interest among Japanese

banks in gaining access to the model is growing, though it did

not clarify whether any have formally sought access.

* The working group will discuss procedures when

vulnerabilities are found, defensive measures and contingency

planning for scenarios where threats cannot be fully contained,

the FSA said. The regulator also said it is considering

information-sharing with U.S. and other overseas authorities.