Japan plans to obtain earlier access to information on undisclosed software vulnerabilities from US AI developers and other overseas technology companies, Nikkei reported and Reuters relayed on November 5. The aim is to shorten the gap between a flaw being found and government agencies or infrastructure operators being able to act on it.
The plan is expected to be part of an emergency cybersecurity package. Japan's government and companies already have access to advanced models from US firms such as Anthropic and OpenAI, and Tokyo recently asked for broader access to help respond to cyberattacks. The main driver is a rise in AI-assisted attacks.
Key details are not yet public: which companies would take part, who would receive pre-disclosure data and how it would be protected until a patch exists. Anthropic's restricted Mythos model, which is strong at finding vulnerabilities, has already drawn requests for access from Japanese megabanks.