The Japanese government and the country’s major financial institutions have secured access to Claude Mythos, an advanced artificial intelligence model developed by U.S.-based Anthropic, in a move that could significantly shore up weaknesses in cybersecurity.

While Finance Minister Satsuki Katayama, who is also the minister of state for financial services, declined Tuesday to disclose how many Japanese financial companies were granted access, three megabanks — MUFG Bank, Mizuho Bank and Sumitomo Mitsui Banking — are reportedly included.

Anthropic announced the same day that it is extending the Mythos partnership to about 150 new organizations in over 15 countries.

Mythos has attracted attention for its ability to identify software vulnerabilities that are unknown even to the service providers themselves. Due to its advanced capabilities, Anthropic had previously stated that the model was too dangerous for public release.

With Mythos, financial institutions will be able to rapidly detect vulnerabilities in their systems and fix them.

Economic crimes of varying degrees occur frequently, and financial institutions are the most vulnerable, said Katayama, who oversees the Financial Services Agency (FSA).

“It is critical that these financial institutions are fully equipped to respond, as it impacts the credibility of Japan’s financial sector,” she said.

Since the financial industry is one of Japan’s key growth sectors, it is imperative that its cybersecurity levels match those of leading U.S. financial companies, Katayama said.

During a meeting last month, U.S. Treasury Secretary Scott Bessent told Katayama that access to Mythos would be granted to the Japanese government and financial companies in a couple of weeks.

Last month, the FSA issued a request to financial institutions, urging them to implement short-term countermeasures under the direct involvement of top management in light of evolving threats posed by frontier AI, or the most advanced AI models.

If a massive number of vulnerabilities are discovered by frontier AI, the FSA urged companies to prioritize tasks and strategically allocate resources where they are needed most, given that the short-term expansion of existing IT department resources is not feasible.

During a meeting with Katayama last week, OpenAI Chief Strategy Officer Jason Kwon promised that the maker of ChatGPT would give some Japanese financial institutions access to GPT-5.5-Cyber, OpenAI’s latest frontier AI model.