OpenAI Provides Japan’s Megabanks with Advanced Cyber Defence Model
May 29, 2026 – 12:08 pm
GPT-5.5-Cyber, OpenAI’s latest model, will be made accessible to MUFG Bank, Sumitomo Mitsui Banking Corporation (SMBC), and Mizuho Bank through their "Trusted Access for Cyber" program. This initiative ensures that only verified defenders gain access to powerful tools designed for cyber defence.
The finance minister, Satsuki Katayama, announced the partnership, emphasizing Japan’s approach to AI as both a threat and a shield. The "Trusted Access" program aims to prevent the misuse of advanced AI models by restricting access to institutions that can be thoroughly vetted.
This collaboration involved direct negotiations between Katayama and US Treasury Secretary Scott Bessent, highlighting its strategic significance. Tokyo is treating cyber defence as a critical national infrastructure, similar to how it procures other strategic capabilities.
The move is part of a broader effort by Japan to address AI-related cyber risks. A public-private working group was established in May, bringing together major banks, the Bank of Japan, and leading AI labs. This group focuses on mitigating vulnerabilities posed by advanced vulnerability-hunting systems, such as Anthropic’s Claude Mythos, which Japanese institutions will also gain access to separately.
The partnership between OpenAI and Japanese megabanks signals the early development of an AI defence contractor market. While it offers enhanced security for large banks, it raises concerns about a potential two-tiered security landscape, leaving smaller banks and fintech startups more vulnerable.