Japan’s three largest banks are preparing to plug OpenAI’s newest AI system into their cyber defense playbooks. Nikkei and Reuters report that MUFG Bank, Sumitomo Mitsui Banking Corp., and Mizuho Bank will gain access to OpenAI’s latest model to help defend against cyberattacks. The model serves only trusted partners and rivals Anthropic’s Claude Mythos in security‑focused capabilities.
Japanese officials say the goal is to use the model to spot potential intrusions and weak points before attackers can exploit them. Finance Minister Satsuki Katayama said officials have authorized “several Japanese financial institutions” to use OpenAI’s G‑5.5‑series model to bolster their defenses against cyber threats. While she did not name specific banks, local media and wire reports point to the three megabanks as the main early users.
Working Alongside Anthropic’s Mythos
OpenAI’s partnership will not be Japan’s only bet on frontier AI for security. Earlier this month, Kyodo and other outlets reported that MUFG, Sumitomo Mitsui, and Mizuho will also gain access to Anthropic’s Claude Mythos, which developers designed to detect and even simulate zero‑day vulnerabilities at scale. Sources told Nikkei that the banks will eventually run both tools side by side, giving their security teams two independent AI “views” of the same systems.
OpenAI plans to provide a specialized GPT‑5.5‑Cyber model to the Japanese government and selected financial institutions, including the three megabanks. Developers tuned the model for cybersecurity tasks, such as analyzing logs, summarizing threat reports, and generating patches or configuration changes.
When paired with Mythos, which focuses on probing code for hidden flaws, the banks hope to shrink the window between a bug appearing and a fix rolling out across their networks.
Tokyo is also building a formal framework around this rapid AI adoption. The Financial Services Agency recently formed a 36-member private working group on AI-driven cyber risk in finance, including MUFG, SMBC, Mizuho, OpenAI, Anthropic, and key cloud providers. The group wants to collaborate on threat intelligence, establish uniform security standards, and assess how much banks can rely on strong models without creating systemic danger.
READ MORE: Algorand Price Breaks Descending Trendline but Stalls Under $0.12
