Nikkei: Japan Asks OpenAI and Anthropic for Wider Access to Advanced Models for Cybersecurity
JUST IN: 馃嚡馃嚨 Japan asks OpenAI and Anthropic for access to their most advanced AI models.
Context
Governments are now asking AI labs for access to their strongest models to find weaknesses in critical systems, which turns frontier models into a matter of national security.
Watcher Guru is a news aggregator account on X. The report it relays comes from Nikkei. Anthropic's Claude Mythos Preview and OpenAI's GPT 5.5 Cyber are models with strong ability to find software vulnerabilities, offered only to limited groups.
Deep dive
Watcher Guru posted on Friday that Japan has asked OpenAI and Anthropic for access to their most advanced AI models. The post cites no source, but the report comes from Nikkei, which published it on Friday.
According to Nikkei Asia, the Japanese government has called on Anthropic and OpenAI to provide greater access to advanced models used to find cybersecurity weaknesses, and will encourage operators of essential infrastructure to use them. The Japanese language Nikkei article says the government wants to find weak points in its important systems as the threat of cyberattacks grows, and will urge power and telecom companies to use the models too.
Nikkei reports that the plan will be part of a cybersecurity package the government is expected to announce as soon as next week, which will include closer work with American AI developers. Japan already has limited access for the government and some financial institutions, according to earlier statements by Japan's financial services minister.
Neither OpenAI nor Anthropic has commented publicly on the request. The access sought is specifically for defensive security uses rather than general access to unreleased models, which is narrower than the Watcher Guru headline suggests.
What to expect next
Watch for the text of Japan's cybersecurity package, expected as soon as next week, and any terms attached to access. OpenAI and Anthropic may announce formal agreements with Japan. Other allied governments could make similar requests, turning access to security focused frontier models into a regular part of diplomacy and national cyber defense. Japan may also seek similar access from Google and other developers of security focused models.