Japan asks Anthropic, OpenAI for wider cybersecurity model access

Tokyo wants U.S. AI developers to share advanced vulnerability-detection tools with critical infrastructure operators as cyberattacks mount. TOKYO — The Japanese government has asked Anthropic and OpenAI to expand access to advanced artificial intelligence models that detect cybersecurity weaknesses, Nikkei has learned. Officials plan to encourage operators of essential infrastructure to adopt these tools. The request comes as Tokyo strengthens national cyber defenses against mounting attacks. Government officials want the U.S. companies to open their most capable security-focused models to Japanese entities that protect power grids, telecommunications networks, and transportation systems. Anthropic and OpenAI both build large language models with specialized functions for analyzing software vulnerabilities and threat patterns. These systems scan code, flag weak points, and generate patch recommendations in minutes rather than days.