Japan asks Anthropic and OpenAI to expand access to advanced cybersecurity models
The Japanese government has asked U.S. artificial intelligence developers Anthropic and OpenAI to provide broader access to advanced models used to find cybersecurity weaknesses, and will encourage operators of critical infrastructure to use these models. According to Nikkei Asia, the move aims to strengthen the country’s cyber defenses in response to ongoing cyberattacks and data breach incidents.
Japan had already obtained some of the related capabilities. Around June 2026, the Japanese government and some financial institutions, including major banks such as Mitsubishi UFJ, Mizuho, and Sumitomo Mitsui, had gained access to Anthropic’s Claude Mythos and OpenAI-related cybersecurity models. The minister for digital affairs and cybersecurity officials previously met with Anthropic co-founder Jack Clark and confirmed that the two sides would cooperate on analyzing AI attack methods. Prime Minister Sanae Takaichi recently asked the Cabinet to advance emergency cybersecurity measures, which are expected to include strengthened cooperation with U.S. AI developers.
The current request focuses on expanding the supply of existing cybersecurity-specific models for critical infrastructure operators to use for vulnerability checks. Nikkei reported that it was not disclosed whether Anthropic and OpenAI had agreed to expand access. The related package of cybersecurity measures is expected to be announced as early as next week. Source: public information.
Original sources:
Title: Japan asks Anthropic, OpenAI for more access to advanced models
Publisher: Nikkei Asia
Publication date: 2026-10-30/31
URL: https://asia.nikkei.com/business/technology/artificial-intelligence/japan-asks-anthropic-openai-for-more-access-to-advanced-models
Nature of source: original reporting (Nikkei has learned)
Verification status: public summary and multiple reprints read
Supporting content: The Japanese government requested expanded access to advanced cybersecurity models, encouraged critical infrastructure operators to use them, and responded to rising attacksTitle: Japan Given Access to Anthropic’s Mythos to Boost Cyber Defense
Publisher: The Japan News / Yomiuri
Publication date: 2026-06-04
URL: https://japannews.yomiuri.co.jp/politics/politics-government/20260604-330887/
Nature of source: previous report
Verification status: checked
Supporting content: The Japanese government and some financial institutions had already gained access to Claude Mythos
ABAB AI Insight
The substance of Japan’s request this time is to further expand the coverage of the already limited access, from the government and major banks that have obtained Mythos and OpenAI cybersecurity models to operators of critical infrastructure, and to incorporate it into the soon-to-be-released national emergency cybersecurity measures. Nikkei’s report emphasizes that the models are intended to find system weaknesses and conduct vulnerability checks, which is a defensive application rather than general model access. The U.S. government had previously imposed access restrictions on some advanced models, and Japan obtained exceptions through coordination with the U.S. Treasury and others, showing a tiered access mechanism within an alliance framework.
For Anthropic and OpenAI, such a request would move cybersecurity-specific models from “exceptions for a few institutions” to a broader set of infrastructure operators, potentially creating a government-backed deployment channel, while also opening room for negotiation over export controls, data security, and misuse risks. Nikkei did not confirm whether the two sides agreed, and actual implementation will depend on model supply conditions, limits on use, and reciprocal arrangements. Japan is facing a rise in corporate data breach incidents, and the prime minister has called for emergency measures to be issued as soon as possible; this request is included in that effort, showing that AI cybersecurity capabilities are being treated as a priority tool for infrastructure defense.
This pattern is consistent with the trend among other allies seeking priority access to frontier models: obtaining controlled models through bilateral negotiations on the grounds of cybersecurity and critical infrastructure protection, rather than broad commercial opening. The actual effect will depend on whether the models are effectively used to fix vulnerabilities, the degree of adoption by operators, and whether more countries make similar requests later. Current public information only shows that Japan has made the request and plans to encourage use; it does not confirm the specific terms or timetable for expanding the supply range.
ABAB News · Cognition Law
Access to cybersecurity-specific models is expanding from exceptions for a few financial institutions to operators of critical infrastructure.
Model access within the alliance framework depends on bilateral negotiations and U.S.-side export control exceptions, rather than automatic opening.
Demand for defensive vulnerability checks has become the main public reason countries request expanded access to advanced models.