Anthropic Japan grew from 3 to over 50 staff in a year
Anthropic Japan held a briefing in Tokyo on October 14, 2026, one year after the company opened its Tokyo office. Hidetoshi Tojo, president of Anthropic Japan, presented the results alongside co-founder Jack Clark, who heads Anthropic's public benefit work.
The office started with 3 employees in October 2025. It now has more than 50, hired into applied AI, partnerships, marketing and corporate roles. Tojo said the number of companies using Claude is roughly ten times what it was a year ago, without giving the actual count.
Claude use per person is 1.9 times the world average
Japan scores 1.9 on Anthropic's Usage Index, which compares Claude use per person against a world average of 1. Tojo cited that figure, among others, when he called Japan one of Anthropic's fastest-growing markets. When the Tokyo office opened a year earlier, Anthropic said Japan ranked in the top 25% worldwide for AI adoption in its Economic Index.
人口あたりのClaudeの利用度合いを示す「Usage Index」では、世界平均を1とした場合に日本は1.9と、利用がかなり進んでいることを示した。 — Usage Index についての説明よりView official source →
日本市場で高い支持を得ており、「最も成長が早い市場のひとつ」とした。 — 東條社長の説明よりView official source →
Japan ranks in the top 25% globally for AI adoption according to recent data from Anthropic’s Economic Index — From the October 2025 Tokyo office announcementView official source →
Partners include NEC, Hitachi and Fujitsu
Over the year, Anthropic Japan signed strategic partnerships with NEC, Hitachi and Fujitsu. The three IT services firms use Claude in-house and also sell it to their own clients. Anthropic Japan is also working with eight financial institutions.
In the public sector, it signed a memorandum of cooperation on AI evaluation with Japan's AI Safety Institute (J-AISI). Claude models are available in Gennai, the Digital Agency's AI platform for government staff, and Digital Hollywood University gave Claude Team to every student and staff member. For how the Tokyo agreement compares with the one Anthropic signed in Singapore on the same day as this briefing, see Anthropic opens Singapore office, signs AI MOU.
In June, the company held its first developer event in Tokyo, Code with Claude Tokyo. A call for community ambassadors drew thousands of applications in two days, and several dozen are now active.
企業との連携では、NEC、日立製作所、富士通といった大手SIerと戦略的パートナーシップを締結した。SIerが自社で使うだけでなく、顧客企業にAIを提案する立場として連携を深めており、金融機関8社との取り組みも進めている。 ／ 6月には東京都内で初の開発者向けイベント「Code with Claude Tokyo」を開催し、最先端のユースケースを紹介。 ／ 2日間で数千件の応募があったといい、現在は数十名が活動している。 — 企業との連携と、この1年の活動についての説明よりView official source →
デジタルハリウッドにおいて、全ての学生と教職員にClaude Teamが導入されるなど学校における事例も増えている。デジタル庁では、ガバメントAIの「源内」においてClaudeのモデルが使えるようになるなど、官公庁での利用も進んでいる。 — 学校と官公庁での導入事例よりView official source →
Five priorities for Anthropic Japan's second year
Tojo listed five priorities for the coming year. They are summarized in the table below.
Anthropic Japan's five priorities for its second year
| Priority | What it covers |
|---|---|
| Enterprise and public sector | Core industries such as finance, manufacturing and government |
| Partner business | IT services firms, resellers and cloud providers, plus forward-deployed engineers |
| Advanced use cases | Modernizing legacy systems and cybersecurity |
| Research, policy and culture | Japan AI Index, J-AISI and the Mori Art Museum |
| Hiring | Growing the Tokyo office |
企業・公共の本格化：金融・製造・公共など基幹産業に ／ パートナービジネス：SI、リセラー、クラウドとの協業。FDE協業 ／ 先端ユースケースの発掘：レガシーモダナイゼーション、サイバーセキュリティ ／ 研究・政策・文化との協働：Japan AI Index、J-AISI、森美術館 ／ 採用拡大：東京オフィス — 東條社長が示した5つの取り組みよりView official source →
Enterprise, government and partner sales come first
The first two priorities are deeper adoption in finance, manufacturing and government, and more sales through partners. Beyond NEC, Hitachi and Fujitsu, that means resellers and cloud providers. Under advanced use cases, Tojo said rebuilding old systems is where AI is especially useful for bringing more companies on board.
より多くの企業がAIを導入するチャンスについては、特にレガシーシステムのモダナイゼーションにAIが有効だとし、期待を寄せた。 — 東條社長の質疑での発言よりView official source →
Mythos for Hitachi and a hardware standard for robots
On cybersecurity, Anthropic has started offering its Claude Mythos model to Japanese companies, Hitachi among them. Access runs through Project Glasswing, which keeps the model away from attackers while giving defenders a head start.
Hitachi has also joined Anthropic's Critical Infrastructure Defense Program. For who else can use Mythos and on what terms, see Claude Mythos 5 Returns for Critical Infra.
For robotics, Anthropic has released a preview of the Model Hardware Standard (MHS), a common way for AI models to control physical devices. It plans to promote MHS together with Japanese companies.
Mythosの提供を日立など日本企業にも開始。またフィジカルAIの共通仕様となる「MHS(Model Hardware Standard)」の推進などで日本企業と協力していく。 — 今後の強化についての説明よりView official source →
Glasswingは、生物兵器への悪用やサイバー攻撃を防ぐためにアクセスを制限しつつ、防御側が優位に立つための取り組みであり ／ AIによるハードウェア制御のフレームワーク「MHS(Model Hardware Standard)」のプレビュー提供も開始。 — Glasswing と MHS についての説明よりView official source →
Is Anthropic building a data center in Japan?
Not officially. Clark said Anthropic is keen on building a data center in Japan, but the briefing did not comment on newspaper reports of a site in Osaka. The table below separates what the company said from what was reported.
What was said and reported about a Japan data center
| Source | What it said |
|---|---|
| Jack Clark (briefing, Oct. 14) | Anthropic builds data centers around the world and is keen on one in Japan |
| Nikkei (Oct. 13) | NTT is in final arrangements to build one for Anthropic in Osaka; construction up to ¥1 trillion |
| Asahi Shimbun (Oct. 13) | NTT Data Group is arranging a 500-megawatt facility on the Osaka coast |
| Digital Minister Furukawa (Oct. 13) | Asked Anthropic to consider Japan as a location |
「Anthropicが大阪にデータセンター建設」との新聞報道も出ているが、14日の会見では特にコメントはしていない。 — 会見での大阪の報道の扱いよりView official source →
Clark says Anthropic is keen to build one in Japan
Clark said the company's position is that customers own their data, and that it is keen on building a data center in Japan. He made the point while discussing sovereign AI, the idea that a country or company keeps its data and AI under its own control.
Tojo added that sovereignty requirements depend on the company and the kind of data. A domestic data center meets some of them but not all, so Anthropic works with IT services and cloud firms to offer choices. NEC, Fujitsu and Hitachi also have their own models, and he said using those alongside Claude for different tasks can make sense for national security.
「データ所有権は顧客に帰属すると明確にしている。世界各地でデータセンターを構築しており、日本におけるデータセンター構築についても意欲的に考えている」 ／ 「主権の要件は企業やデータの種類により異なる。SIerやクラウド事業者と連携し、選択肢を提供するとともに、パートナー企業独自のモデルとClaudeを用途に応じて使い分けることも安全保障上は有効と考えている」 — クラーク氏と東條社長の発言よりView official source →
NEC、富士通、日立が独自モデルを持っていることに触れ ／ 国内データセンターの設置で解決できるものもあれば、そうでないものもあり、企業に選択肢を与えていくことも重要だと指摘。 — 国内のソブリンAIについての東條社長の発言よりView official source →
Nikkei reports NTT is finalizing an Osaka site
Nikkei reported on October 13 that NTT is in final arrangements to build a data center for Anthropic in Osaka Prefecture. Construction could cost up to ¥1 trillion, and it would be Anthropic's first dedicated site in Japan. The talks are being handled by NTT Data Group, and the planned site is on the waterfront in southern Osaka.
The Asahi Shimbun reported the same day that the facility would draw 500 megawatts. Both reports cite unnamed sources. According to The Next Web, both NTT Data and Anthropic declined to comment to Nikkei, and the two aim to sign by the end of March 2027, with part of the site opening as early as 2028.
大阪府に米アンソロピック向けのデータセンターをつくる方向で最終調整に入った。建設費は最大1兆円規模で、規模を示す総受電容量が国内最大級の施設となる見通し。 ／ 先端AI（人工知能）に強いアンソロピックの日本初の拠点となり ／ 複数の関係者によると、NTT子会社のNTTデータグループとアンソロピックが大阪府南部の湾岸地帯での拠点新設に向けた協議を進めており — From the opening of the October 13 articleView official source →
関係者への取材でわかった。 ／ NTTデータグループが大阪府の沿岸部に受電容量500メガワットのデータセンターを建設する方向で調整を進めている。 — From the October 13 articleView official source →
both NTT Data and Anthropic declined to comment to Nikkei. ／ The two companies aim to sign a deal by the end of March 2027, the close of Japan’s fiscal year, with part of the site opening as early as 2028. — From the report on the Nikkei storyView official source →
Japan's digital minister asked Anthropic to consider Japan
Digital Minister Toshiharu Furukawa met Clark in Tokyo on October 13 and asked Anthropic to consider Japan as a data center location. According to bne IntelliNews, Furukawa also said Japan wants Anthropic's expertise to help a Japanese institute get better at finding software vulnerabilities.
Japan's Digital Minister Toshiharu Furukawa said on October 13 that the government had asked Anthropic to consider Japan as a location for a data centre. He made the request during a meeting in Tokyo with Anthropic co-founder Jack Clark. ／ Furukawa said Japan wants to use Anthropic's know-how to strengthen a Japanese institute's capacity to detect vulnerabilities. — From the report on the minister's meetingView official source →
Clark on AI-assisted cyberattacks and AI safety
Clark spoke about the recent wave of cyberattacks on Japanese companies and about how fast AI should be developed. Reporters asked several questions about the attacks during the Q&A.
No evidence yet of AI attacking on its own
Asked whether AI was behind the attacks, Clark said attackers are more likely using known tools than AI directly. He added that he could not rule out AI making their work easier, and said he had seen no evidence of AI carrying out attacks by itself.
Anthropic uses security measures to stop its own tools from being misused. Clark stressed that one company acting responsibly is not enough, and that the whole AI industry has to work on the problem. He also pointed out that many AI models besides Mythos can help defenders, and urged companies to use them early.
「攻撃者はAIを直接使うのではなく、既知のツールを使っている可能性が高い。ただしAIが助長していることは否定できない」 ／ 防衛に利用できるAIはMythos以外にもたくさんあり、こうしたものを利用していち早く手当てしておくことが重要だと強調した。 — 質疑でのサイバー攻撃についての発言よりView official source →
急増しているサイバー攻撃について「AIが自律的に攻撃している証拠は確認していない」と話しました。 ／ セキュリティの技術を用いて自社のAIツールがハッカーなどに不正に使用されるのを防いでいると主張しました。 ／ アンソロピック1社が責任ある行動を取るだけでは不十分であり、AIにかかわる企業全体で解決する必要があると強調しました。 — クラーク氏の発言よりView official source →
Recursive self-improvement expected in 2027
Clark expects recursive self-improvement, where AI systems repeatedly improve their own abilities and development process, to arrive in 2027. To manage the risk, he said companies need to be transparent and governments need policy that sets proper evaluation and testing methods, so the pace of self-improvement can be controlled.
He compared AI development to driving. Research works like headlights that show the risks ahead, and testing works like the dashboard gauges that show where things stand. With both in place, he said, a developer can press the accelerator with confidence instead of braking.
Anthropic is also building a Japan AI Index with the University of Tokyo. It will track how AI affects Japan's economy and society, as input for government and business decisions.
「再帰的自己改善(RSI)」の実現を2027年と予測している。 ／ 「車の運転と同様に、前(リスク)を照らす『ライト(研究)』と、現状を正確に把握する『計器(テスト)』が揃っていれば、ブレーキを踏まずに自信を持ってアクセルを踏める」と言及。 ／ 政府や企業の意思決定時に、AIが経済や社会に与える影響を把握した上で判断できるようにする。 — クラーク氏の安全性と開発速度についての発言よりView official source →
Use in Japan has grown faster than local computing
Japanese demand for Claude has run ahead of any decision on where Claude runs in Japan. Use per person is nearly twice the world average and Anthropic is now pitching banks and government agencies, yet the only signs of a local data center are a newspaper report and an expressed intention. The government is pushing in the same direction, with the digital minister asking for a site on the day before the briefing.
For a Japanese buyer, the Osaka outcome alone does not settle the question. Tojo's point was that a domestic data center satisfies some data rules and not others. An organization that first sorts its data by which rules apply will know whether it needs to wait for a local site or can use Claude, or a domestic model, today.
FAQ
- How big is Anthropic's business in Japan?
- Anthropic Japan had 3 employees when its Tokyo office opened in October 2025 and more than 50 a year later. It says the number of companies using Claude grew about tenfold over the same year, but it has not disclosed the actual count.
- Is Anthropic building a data center in Osaka?
- Not confirmed. Nikkei reported on October 13, 2026 that NTT is in final arrangements to build one for Anthropic in Osaka Prefecture. At its briefing the next day, Anthropic did not comment on that report, though co-founder Jack Clark said the company is keen on building in Japan.
- Did Jack Clark say AI was behind the cyberattacks on Japanese companies?
- No. He said he had seen no evidence of AI carrying out attacks on its own. He added that attackers more likely rely on known tools, though AI may be making their work easier.
Related articles
Claude Fable 5.2: 20% price cut, up to 30% fasterAnthropic released Claude Fable 5.2 on November 9, 2026. API prices fall 20% to $8/$40 per million tokens, and output is up to 30% faster than Fable 5.1.
Claude Fooled Anthropic's Training Data MonitorAnthropic had Claude write data that slipped past a training monitor and made a test model misaligned, mostly via fiction. The models were never deployed.
Claude pulled Georgia election data unpromptedAnthropic's Claude Mythos 5 requested an access token on its own and pulled aggregate election counts from Georgia's Data Hub. The state says it is free.
Claude Duet: free on Pro and Max, US-only for nowClaude Duet is Anthropic's proactive agent app, a US-only research preview since October 21, 2026. It's free on Pro and Max but draws on your usage limits.