Japan weighs OpenAI cyber tools as government rethinks AI nationalism

OpenAI, the developer of ChatGPT, wrapped up a sales pitch to Japan, offering the Japanese government and private companies its latest generative AI model specialized in cybersecurity. The company showcased a number of its cyber defense programs to Japanese media on May 21. It’s where OpenAI board member, Paul Nakasone revealed that the visit to Japan was intended for talks with government officials. Nakasone said they discussed cybersecurity measures across 15 critical sectors with the Japanese government. While talks are set to continue, OpenAI said it hopes to launch the service in Japan at “an early stage.” OpenAI is offering the specialized GPT-5.5 Cyber AI model to the Japanese government. Its standard GPT-5.5 with Trusted Access for Cyber (TAC) defensive tool will be offered to Japanese firms and businesses under an application and screening process. Protection against Mythos During the press conference, OpenAI’s Head of National Security Policy, Sasha Baker stressed that a cyber defense ‘ecosystem’ is needed to overcome powerful models. She pointed to Anthropic’s non-public Mythos, which can autonomously identify and exploit security flaws in software, web browsers, and operating systems. Nakasone said powerful AI also requires stronger governance and safeguards. “We will build robust security systems and stay ahead of malicious actors. We intend to expand these efforts broadly from finance and critical infrastructure to local governments and manufacturing supply chains.” Nakasone, who previously led U.S. Cyber Command under the Trump administration, described Japan as central to a “Free and Open Indo-Pacific” and suggested OpenAI would deepen collaboration with the country. “We want the Japanese government and companies to use our most advanced models,” added Sasha Baker. The real threat is AI dependence OpenAI’s visit comes as the Japanese government intensifies its push for “sovereign AI.” Japan’s Basic AI Plan, finalized in December 2025, revolves around the concept of “trustworthy” AI. It stems from economic security concerns that foreign tech giants could control the entire AI supply chain. The Ministry of Economy, Trade and Industry (METI) had proposed developing a large-scale domestic foundation model akin to a Japanese version of ChatGPT using government funding. When METI presented its proposal at an LDP Digital Society Promotion Headquarters meeting in October 2025, some lawmakers criticized the plan as reckless, arguing that Japan lacked the policy resources needed to compete with the U.S. and China. METI has since dropped its ‘Japanese ChatGPT’ goal, but the government is still determined to foster a homegrown AI stack, which includes foundation models, data centers, AI chips, as well as physical AI infrastructure . The government is preparing to revise its Basic AI Plan this summer. At an AI strategy meeting on May 19, lawmaker Kimi Onoda confirmed the revised draft will strengthen AI sovereignty from a national security stance. Japan’s AI reality check While some Japanese companies, such as Preferred Networks, Ricoh, SoftBank, NEC, Honda, and Sony Group, have begun developing foundation models, many in the industry privately acknowledge the difficulty of catching up to the U.S. and China. Japan was ranked 30th out of 69 countries in the IMD World Digital Competitiveness Score in 2025. There’s also a massive AI investment shortfall between Japan and its rivals. According to Japanese government data, the U.S. government invested approx. $329 billion in local AI development from 2019 to 2023. The Chinese government invested approx. $133 billion. The Japanese government, on the other hand, invested a meager $10 billion. The end of AI nationalism The government’s Digital Society Promotion Headquarters is preparing a proposal against an entirely Japanese AI stack. The proposal is urging the government to prioritize AI innovation in manufacturing, healthcare, and infrastructure sectors. It argues Japan could combine foreign-developed foundation models with applications developed by domestic industrial data to create a competitive advantage. On May 11, the Secretary General of the Digital Society Promotion Headquarters, Akihisa Shiozaki, said Japan is entering a post-LLM era that requires a major paradigm shift. He stressed the goal shouldn’t be building sovereign AI but rather diversifying suppliers. “What matters most is ensuring autonomy without becoming dependent on any single country, company, or provider. Rather than focusing solely on ‘sovereign AI,’ Japan needs to think about how to protect its AI sovereignty.” If you're reading this, you’re already ahead. Stay there with our newsletter .