On May 29, OpenAI released the “Japan Cyber Action Plan” and simultaneously granted access to the cybersecurity-specific GPT-5.5 Cyber model, which had been launched on May 7, to Japan. The first users include Mitsubishi UFJ Bank, Sumitomo Mitsui Banking Corporation, and Mizuho Financial Group, with plans to expand beyond financial institutions. On the same day, OpenAI also signed a memorandum of cooperation with Japan’s AI Safety Institute (AISI), which is responsible for AI safety evaluation under the Japanese government. This marks OpenAI’s third such arrangement, following similar collaborations with the United States and the United Kingdom, and its first in Asia. The two parties will cooperate on risk verification for next-generation models.
According to Nikkei Chinese Edition, this move follows Anthropic’s steps as OpenAI’s latest effort to build a joint Japan-U.S. government and private sector cooperative mechanism to counter cyberattacks. The two companies have recently engaged in consultations with allied governments, and cybersecurity is becoming the new main battleground for AI competition.