Nikkei Asia reported on 30 October that “The Japanese government has called on U.S. artificial intelligence developers Anthropic and OpenAI to provide greater access to advanced models used to find cybersecurity weaknesses and will encourage their use by operators of essential infrastructure, Nikkei has learned.”
By our arithmetic, the request comes 171 days after Kyodo reported, on 12 May, that Japan was negotiating with Anthropic for access to its latest model. Taking the Japan Times’ 3 June report of Japan’s Mythos access to refer to Tuesday 2 June, that access came about 10 days before a U.S. order on 12 June, by our arithmetic.
Business Pill · Sovereign AI
A short explainer of sovereign AI: why governments want their own access to frontier models and compute. It teaches the general idea only and says nothing about this request.
The key insight: Nikkei Asia reports that Japan has asked Anthropic and OpenAI for greater access to advanced models used to find cybersecurity weaknesses, 171 days after Kyodo reported Japan’s talks with Anthropic, by our arithmetic.
Nikkei’s article, by Kanako Chiba, is paywalled, and only its headline, subheading and first paragraph were readable. The subheading reads “Tokyo moves to bolster nation’s cyber defenses as attacks mount”.
A relay of the same report on X by @DeItaone added: “JAPANESE GOVT PLANS TO INCLUDE STRONGER COOPERATION WITH U.S. AI DEVELOPERS IN CYBERSECURITY PACKAGE DUE OUT AS EARLY AS NEXT WEEK – NIKKEI”.
Kyodo reported on 26 October that Prime Minister Sanae Takaichi “said Monday the government will compile by the end of next week an emergency package of additional cybersecurity measures, following a spate of data breaches at companies”.
Calendar days between dated reports in Japan’s 2026 push for access to frontier cyber models: Kyodo’s 12 May report of talks with Anthropic, the Japan Times’ 3 June report of access (taken as 2 June), the 12 June U.S. order, CNBC’s 30 June report of restored access for some U.S. organizations, and Nikkei Asia’s 30 October report. Our arithmetic.
What Was Posted on X
JAPANESE GOVT PLANS TO INCLUDE STRONGER COOPERATION WITH U.S. AI DEVELOPERS IN CYBERSECURITY PACKAGE DUE OUT AS EARLY AS NEXT WEEK – NIKKEI
@DeItaone relayed Nikkei’s report that the Japanese government plans to include stronger cooperation with U.S. AI developers in a cybersecurity package due out as early as next week.
How Japan Got Access in June
Kyodo reported on 12 May that “The Japanese government is negotiating with U.S. artificial intelligence company Anthropic to obtain access to its latest AI model, government sources said Tuesday, amid growing concerns over cyberattacks using sophisticated AI tools”.
On 21 May, Kyodo reported that OpenAI “said Thursday it will provide the Japanese government and some companies with an AI model featuring advanced cybersecurity capabilities”.
The Japan Times reported on 3 June that “The Japanese government and the countryโs major financial institutions have secured access to Claude Mythos”, and that MUFG Bank, Mizuho Bank and Sumitomo Mitsui Banking “are reportedly included”. It also reported that OpenAI’s Jason Kwon promised some Japanese financial institutions access to “GPT-5.5-Cyber, OpenAIโs latest frontier AI model”.
The 12 June U.S. Order
The Japan Times reported on 14 June that “Anthropic has disabled access to its most advanced artificial intelligence models, including Mythos, following an unprecedented order by the Trump administration to keep the technology out of the hands of all foreign nationals”. We covered the order at the time in our 19 June piece.
CNBC reported on 30 June that “Anthropic also said it has restored access to Mythos 5 for some U.S. organizations, following government approval granted”, and that it would “continue to work with the U.S. government to expand access to Mythos 5 for more domestic and international partners through its Glasswing program”.
Other Limits on Access in 2026
Politico reported on 24 September that “The White House has asked OpenAI and Anthropic not to share their new artificial intelligence models with the U.K. governmentโs testing agency until the models have gone through testing with the U.S. government”, according to a person familiar with the matter and a senior U.S. administration official.
Politico added that Anthropic did not provide Claude Mythos 5.1 to the UK’s AI Security Institute, saying the model was “only available to a set of U.S. organizations”. An Anthropic blog quoted by Politico reads “Weโre coordinating with the U.S. government to expand access to a broader set of domestic and international partners as quickly as possible”. We covered Mythos 5.1’s three cyber tiers on 7 October.
By our arithmetic, 12 May to 30 October is 171 days. Taking the Japan Times’ 3 June report to refer to Tuesday 2 June, and the U.S. order to 12 June, the date CNBC gives, the reported access came about 10 days before the order. CNBC’s report of restored access came 18 days after the order, and 122 days before Nikkei’s report.
The Structural Read
As we read the reports, Japan’s push spans both U.S. labs: Kyodo reported talks with Anthropic in May and an OpenAI offer to the government, and Nikkei now names both companies.
As we read the June reports, Japan’s government and megabanks were reported to have Mythos access in early June, and a U.S. order on 12 June covered all foreign nationals.
As we read Politico’s September report, access to new models for at least one foreign government tester has also run through U.S. review.
Nikkei Asia subheading, 30 October 2026
“Tokyo moves to bolster nation’s cyber defenses as attacks mount”
Three Implications
FOR ANTHROPIC AND OPENAI Nikkei reports Japan wants greater access to their advanced cyber models; Nikkei’s readable text does not say whether either company has agreed.
FOR JAPAN’S INFRASTRUCTURE OPERATORS Nikkei reports the government will encourage operators of essential infrastructure to use such models.
FOR JAPAN’S CYBER PACKAGE Kyodo reports Takaichi said an emergency package would be compiled by the end of next week, and @DeItaone’s relay of Nikkei says stronger cooperation with U.S. AI developers is planned for it.
The framework says “It is where decisions are made about who gets access to frontier intelligence, under what conditions, and for which purposes.”
As we read it, Japan’s request is a government asking for a wider place in that access decision. The framework was published in June 2026 and does not discuss this request.
What Is Not Established
Nikkei’s readable text does not say which models Japan asked for, on what terms, or whether Anthropic or OpenAI has agreed. We found no statement from either company on the request as of publication. Whether U.S. approval would be needed for wider access is not stated in the sources we read.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Nikkei Asia reports that Japan has asked Anthropic and OpenAI for greater access to advanced models that find cybersecurity weaknesses, ahead of an emergency cyber package Takaichi said would come by the end of next week. By our arithmetic, it comes 171 days after Kyodo reported Japan’s talks with Anthropic, and about 10 days separated Japan’s reported Mythos access in June from the U.S. order.
A note on sourcing. The report is Nikkei Asia’s; its article is paywalled and only the headline, subheading and first paragraph were readable. Earlier events come from Kyodo, The Japan Times, CNBC and Politico as cited. Day counts are our own arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
