Japan’s three largest banks are set to gain access to Anthropic’s advanced artificial intelligence model, Mythos, in about two weeks, according to a source. This development comes amid growing concerns from cybersecurity experts who warn that Mythos can rapidly uncover and exploit vulnerabilities in legacy banking IT systems, prompting regulatory alerts. Finance Minister Satsuki Katayama has recently taken action by establishing a public-private working group to address these cybersecurity risks, following a meeting with U.S. Treasury Secretary Scott Bessent. This group will hold its first meeting on Thursday, underscoring the urgency of the situation as U.S. banks have already begun testing Mythos.
Anthropic: Anthropic is a U.S. AI research firm focused on building safe and interpretable AI systems, with its latest Claude Mythos model demonstrating unprecedented capabilities in detecting and exploiting IT vulnerabilities. The company has granted Mythos access to U.S. banks for defensive testing and aims to expand to Japanese megabanks and European institutions amid global cybersecurity concerns. Anthropic did not respond to requests for comment on the Japan rollout.
Scott Bessent: Scott Bessent is the U.S. Treasury Secretary, overseeing fiscal policy, international finance, and economic security under President Trump. He recently engaged with Japan’s Finance Minister Satsuki Katayama in Tokyo on topics including AI cybersecurity threats like Mythos impacting financial systems. The discussions preceded Japan’s decision to form a working group on these risks.
Satsuki Katayama: Satsuki Katayama serves as Japan’s Finance Minister and Minister for Financial Services, leading policy on financial stability and regulation. She met U.S. Treasury Secretary Scott Bessent and announced a public-private working group to tackle cybersecurity risks from Anthropic’s Mythos to the banking sector. The group’s first meeting is scheduled for Thursday following her directives.
Mizuho Financial Group: Mizuho Financial Group is a major Japanese banking conglomerate providing comprehensive financial services including commercial banking, securities, and asset management worldwide. Along with other megabanks, it is set to receive access to Anthropic’s Mythos AI model soon to address vulnerabilities in legacy systems highlighted by cybersecurity experts. The group has been involved in recent talks with Japanese regulators on AI-related financial risks.
Mitsubishi UFJ Financial Group: Mitsubishi UFJ Financial Group is one of Japan’s largest financial institutions, operating extensive retail, corporate, and investment banking services globally through subsidiaries like MUFG Bank. It is expected to gain access to Anthropic’s Mythos AI model in about two weeks to test and bolster cybersecurity defenses against advanced AI threats. The bank declined to comment on the development amid regulatory discussions.
Sumitomo Mitsui Financial Group: Sumitomo Mitsui Financial Group operates as one of Japan’s top megabanks, offering a broad spectrum of banking, trust, and securities services internationally via entities like Sumitomo Mitsui Banking Corp. It is among the three largest banks slated for Mythos AI access in the coming weeks to mitigate cybersecurity challenges posed by the model. Executives participated in preparatory discussions with officials on AI cybersecurity measures.
Japan Response: Finance Minister Katayama initiated a public-private group post-Bessent meeting to develop strategies against Mythos cybersecurity threats.
Global Bank Access: U.S. banks have already tested Mythos defensively, with expansion planned for European and UK banks after Japan’s megabanks.
Mythos Capabilities: Anthropic’s Claude Mythos model rapidly uncovers and exploits flaws in legacy banking IT systems, spurring global warnings from regulators and experts.