Why This Topic Landed in Japan

Chinese AI models such as Moonshot's Kimi and DeepSeek had been promoted as rivals to the most advanced US systems, and Japanese users had followed the "China AI threat" narrative closely. Anthropic's early-September 2026 threat-intelligence report described Moonshot and DeepSeek forwarding some user requests to Claude and returning Claude's answers as their own, using them for distillation. To many readers the report read as a sharp reversal — evidence that the "homegrown" models were a relabeled relay. The report also noted that sensitive inputs, including Chinese surveillance data and Russian government-linked credentials, may have been routed to Claude; the wider claim that "military secrets" leaked wholesale, along with reports of a stock crash and detained developers, remains unverified.

Key Reaction Themes

  • "The 'homegrown' model was just a relabel" (most common) — Most comments mocked the models as ripped-off shells, recalling an earlier case where a Chinese model turned out to be ChatGPT inside, and asking what the "China AI threat" had ever amounted to.
  • Alarm at the security angle (next largest) — A large group fixed on the irony of officials and military-linked users typing secrets into what they believed was a domestic AI, only for the data to pass through a rival country's servers.
  • Technical and cynical asides (a smaller group) — Some distinguished distillation from simple API relaying, or noted that plenty of ordinary users must have typed corporate secrets in as well.

What Japanese Netizens Are Saying

China's AI models accused of routing queries to Claude

A report that "homegrown" Chinese models were relaying user input to Claude drew a wave of mockery and unease over leaked data.

CNET Japan
CNET Japan
@cnet_japan
·

中国AI「Kimi」、利用者の入力をClaudeに転送し、Claudeの回答を利用者に表示か--Anthropicが報告 japan.cnet.com/article/352525…

japan.cnet.com中国AI「Kimi」、利用者の入力をClaudeに転送し、Claudeの回答を利用者に表示か--Anthropicが報告Anthropicは、AIの悪用事例をまとめた2026年9月版の脅威レポートを公開した。DeepSeekや、「Kimi」を開発するMoonshot AIが、利用者の入力をClaudeに転送し、その回答を利用者に返していたと報告した。
Reply
Read 70 replies
Anthropic
Anthropic
@AnthropicAI
·

We're publishing our most detailed threat intelligence report to date. It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them. We disrupted every operation in the report, Show more

www.anthropic.comCountering misuse of AI: September 2026 / AnthropicCase studies from threat actors disrupted between December 2025 and August 2026 across seven areas of harm, from cyber operations to biological misuse.
Reply
Read 2.9K replies
kokko | 古賀興司
kokko | 古賀興司
@kokko_coco
·

中国AIのKimiがClaudeを流用したのが流れてるが、Anthoropicのレポートを見るともっと衝撃的だった。人民解放軍関係者の可能性が高いユーザーが、監視カメラデータをKimiに読み込ませ、ある人物の行動を分析しようとしたらしい。それで、KimiはこのデータをClaudeへ転送した。結果、この監視データの画 Show more

Image
Reply
Read 73 replies

Sources: CNET Japan, 2ch

Comments:

  • "Told you it was all ripped off."
  • "The one that made news before was the same — just a new shell with ChatGPT inside."
  • "So what was all that 'China AI threat' talk about?"
  • "If it's Claude inside, how is it 'beating' Claude's performance?"
  • "Simple questions go to Kimi, complex reasoning gets dumped on Claude — and military people, not knowing that, typed secrets into the AI and leaked them."
  • "It means their political and military plans were laid bare — Xi and Putin are finished."
  • "Even setting aside state secrets, there must be quite a few fools who typed corporate secrets in too."
  • "Chinese AI stocks are crashing lol"
  • "How kind of you, America~" [Translator's note: sarcastic.]
  • "There was a similar story in the first half of this year too — how many times are people going to get fooled?"