Anthropic's Claude Mythos Security AI Set for Japanese Megabank Deployment via Project Glasswing

Three of Japan's largest financial institutions are on the verge of becoming the first non-Anthropic entities to gain access to Claude Mythos, the company's vulnerability-hunting AI model, according to sources familiar with the matter. MUFG, Mizuho, and SMFG are expected to receive access to the system within roughly two weeks, joining the restricted rollout of Anthropic's Project Glasswing initiative.

Project Glasswing represents Anthropic's structured approach to deploying its AI capabilities for security-focused applications. Claude Mythos specifically functions as a vulnerability detection tool, designed to identify weaknesses in systems and code. The addition of Japan's megabanks marks a significant expansion of the project's participant base, moving beyond Anthropic's internal operations into external institutional deployment. The timing of the rollout comes as financial institutions worldwide face mounting pressure to strengthen their cybersecurity infrastructure against increasingly sophisticated threats.

The deployment raises questions about how major financial institutions will integrate AI-driven vulnerability detection into existing security frameworks. Japanese banks, which manage trillions of dollars in assets and serve millions of customers, represent high-value targets for cyber threat actors. Industry observers note that the adoption of advanced AI security tools by such prominent financial institutions could signal a broader shift in how the sector approaches threat detection and system hardening. The source characterized the development as imminent, though neither Anthropic nor the banks have issued official statements confirming timelines or specific use cases.