Japan’s government and some of the country’s biggest banks are set to gain access to Anthropic’s Claude Mythos under a cybersecurity initiative discussed by Finance Minister Satsuki Katayama.

She said a public-private working group had been created to examine risks the model could pose to the financial system.

MUFG Bank, Mizuho Bank, and Sumitomo Mitsui Banking Corp. were among the lenders expected to receive access.

Anthropic said Mythos Preview is an unreleased model for defensive cybersecurity work such as local vulnerability detection and penetration testing.

The company has found thousands of high-severity flaws, including zero-days, across major operating systems and web browsers.

Katayama said several Japanese financial groups had been granted access to OpenAI’s GPT-5.5 for cyber defense under restricted-access arrangements.

The move suggests OpenAI and its partners are positioning advanced coding models as tools to strengthen cybersecurity in the financial sector.

🔗 Source: The Japan Times

Recent Anthropic developments