China Finds Security Problems in Claude Code
China Finds Security Problems in Claude Code
中國發現 Claude Code 存在安全問題
Introduction
The Chinese government says the AI tool Claude Code is not safe.
中國政府表示 AI 工具 Claude Code 並不安全。
Main Body
A Chinese group found a problem in some versions of the software. The tool sends user names and locations to other computers. The government says this is a big danger.
一個中國團體在部分版本的軟體中發現了一個問題。該工具會將使用者名稱和位置傳送到其他電腦。政府表示這是一個巨大的危險。
Anthropic is the company that made the tool. An engineer said the code was a test. He wanted to stop people from stealing the AI.
Anthropic 是開發該工具的公司。一名工程師表示該程式碼僅為測試,目的是為了防止他人盜用該 AI。
Now, many companies in China are stopping their work with Claude Code. For example, Alibaba told its workers to stop using it on July 10.
現在,許多中國公司正停止使用 Claude Code。例如,阿里巴巴於 7 月 10 日要求員工停止使用。
Conclusion
China tells people to delete the old software or get a new version now.
中國要求民眾立即刪除舊版軟體或更新至新版本。
Vocabulary Learning
⚡ The 'Action' Pattern
Look at how we describe things happening right now or in a fixed way in this text. This is the key to A2 level speaking.
1. The 'S' Rule (Present Simple) When one person or thing does something, add an -s to the action word:
- The government says... (It says)
- The tool sends... (It sends)
2. The 'ING' Rule (Happening Now) When something is in progress, we use -ing:
- ...stopping their work (They are doing this now)
- ...using it (They are doing this now)
3. Simple Words for Big Ideas Instead of hard words, use these A2 basics found in the text:
- Problem Something wrong.
- Danger Not safe.
- Version A type of software.
Quick Summary:
Person/Thing + Action-s = Fact.
Person/Thing + Action-ing = Process.
Vocabulary Learning
Chinese Authorities Find Security Risks in Anthropic's Claude Code
中國當局發現 Anthropic 的 Claude Code 存在安全風險
Introduction
The Chinese government has issued an official warning about security risks related to Claude Code, an AI coding tool created by the American company Anthropic.
中國政府已針對美國公司 Anthropic 開發的 AI 編碼工具 Claude Code 的安全風險發出官方警告。
Main Body
The National Vulnerability Database (NVDB), which works with the Ministry of Industry and Information Technology, reported a security 'backdoor' in certain versions of Claude Code. Specifically, versions 2.1.91 to 2.1.196, released between April and June, can send user identity and location data to remote servers without the user's permission. The regulator described this as a serious threat; consequently, they advised organizations to monitor their network traffic and either delete the software or update to version 2.1.204.
與工業和信息化部合作的國家漏洞數據庫 (NVDB) 報告指出,Claude Code 的某些版本存在安全「後門」。具體而言,在 4 月至 6 月期間發布的 2.1.91 到 2.1.196 版本,可以在未經用戶許可的情況下,將用戶身份與位置數據傳送到遠端伺服器。監管機構將此描述為嚴重威脅;因此,他們建議機構監控網路流量,並刪除該軟體或更新至 2.1.204 版本。
This situation happens during a time of high tension regarding national security and intellectual property. Although Anthropic restricts users in China, many people still access the tool using VPNs. Furthermore, Anthropic previously claimed that Chinese companies, such as Alibaba, used 'distillation' to copy their AI models. In response, Anthropic engineer Thariq Shihipar explained that the problematic code was an experiment started in March to stop unauthorized resellers and prevent this copying process. He emphasized that the feature was planned for removal in a later update.
此情況發生在國家安全與知識產權高度緊張的時期。雖然 Anthropic 限制中國用戶使用,但許多人仍透過 VPN 存取該工具。此外,Anthropic 先前聲稱阿里巴巴等中國公司利用「蒸餾」技術來抄襲其 AI 模型。對此,Anthropic 工程師 Thariq Shihipar 解釋,有問題的代碼是 3 月開始的一項實驗,旨在阻止未經授權的轉售商並防止此抄襲過程。他強調,該功能計劃在隨後的更新中移除。
Chinese companies have reacted quickly to these warnings. For example, reports show that Alibaba ordered its employees to stop using Claude Code for professional work starting July 10, citing security concerns.
中國公司對這些警告反應迅速。例如,報告顯示阿里巴巴由 7 月 10 日起,以安全顧慮為由,要求員工停止將 Claude Code 用於專業工作。
Conclusion
Chinese regulators have urged users to fix their Claude Code installations immediately after discovering that the software could transmit data without authorization.
中國監管機構在發現該軟體可能會在未經授權下傳輸數據後,促請用戶立即修復 Claude Code 的安裝。
Vocabulary Learning
🚀 The 'Logic Bridge': Moving from A2 to B2
At the A2 level, you connect ideas with simple words like and, but, or because. To reach B2, you need Logical Connectors. These words tell the reader how two ideas relate (cause, contrast, or addition) without using the same basic words over and over.
🔍 Analysis of the Text
Look at how the article moves from one idea to the next. Instead of basic words, it uses "Bridge Words":
-
The Result Bridge: Consequently
- A2 style: "It was a threat, so they told companies to delete it."
- B2 style: "The regulator described this as a serious threat; consequently, they advised organizations to... delete the software."
- Why it works: Consequently is a formal way to say "as a result." It signals a professional, academic tone.
-
The Contrast Bridge: Although
- A2 style: "Anthropic blocks China, but people use VPNs."
- B2 style: "Although Anthropic restricts users in China, many people still access the tool..."
- Why it works: Although allows you to put two opposing ideas in one elegant sentence. It shows you can handle complex sentence structures.
-
The Addition Bridge: Furthermore
- A2 style: "And Anthropic said Alibaba copied them."
- B2 style: "Furthermore, Anthropic previously claimed that Chinese companies... used distillation."
- Why it works: Furthermore is used when you are adding a second, more important point to your argument.
🛠️ Quick Upgrade Guide
| If you want to say... | Stop using (A2) | Start using (B2) |
|---|---|---|
| "And also" | And / Also | Furthermore / Moreover |
| "But" | But / However | Although / Despite this |
| "So" | So | Consequently / Therefore |
Pro Tip: Notice that Consequently and Furthermore are often followed by a comma (,) when they start a sentence. This is a key punctuation mark for B2 writing!
Vocabulary Learning
Identification of Security Vulnerabilities in Anthropic's Claude Code by Chinese Regulatory Authorities
中國監管機構在 Anthropic 的 Claude Code 中發現安全漏洞
Introduction
The Chinese government has issued a formal warning regarding security risks associated with the AI coding tool Claude Code, developed by the U.S.-based firm Anthropic.
中國政府已就美國公司 Anthropic 開發的 AI 程式碼工具 Claude Code 的安全風險發出正式警告。
Main Body
The National Vulnerability Database (NVDB), an entity affiliated with the Ministry of Industry and Information Technology, reported the detection of a security backdoor within specific iterations of Claude Code. According to the NVDB, versions 2.1.91 through 2.1.196—released between April 2 and June 29—possess a mechanism capable of transmitting user identity and geolocation data to remote servers without explicit consent. The regulator characterized this functionality as a severe threat, subsequently advising institutions to implement rigorous network traffic monitoring and to either uninstall the software or migrate to version 2.1.204.
國家漏洞資料庫(NVDB)是工業和信息化部下屬的一個機構,該機構報告在 Claude Code 的特定版本中偵測到安全後門。根據 NVDB 的說法,在 4 月 2 日至 6 月 29 日之間發佈的 2.1.91 至 2.1.196 版本,擁有一套能在未經明確同意的情況下,將使用者身分與地理位置數據傳送至遠端伺服器的機制。監管機構將此功能定性為嚴重威脅,隨後建議各機構實施嚴格的網路流量監控,並採取解除安裝軟體或遷移至 2.1.204 版本的措施。
This development occurs within a context of systemic friction regarding intellectual property and national security. Anthropic has maintained restrictions on users in China and other designated adversarial jurisdictions; however, access has persisted via third-party proxies and virtual private networks. Furthermore, Anthropic has previously alleged that Chinese entities, specifically Alibaba, engaged in 'distillation'—the process of reverse-engineering AI models to replicate their capabilities. In response to these findings, an Anthropic engineer, Thariq Shihipar, stated that the disputed code was an experimental measure initiated in March to mitigate account abuse by unauthorized resellers and to prevent the aforementioned distillation. Shihipar indicated that the functionality was slated for removal in a subsequent release.
此發展發生在關於知識產權與國家安全存在系統性摩擦的背景之下。Anthropic 一直對中國及其他指定對手司法管轄區的使用者維持限制;然而,使用者仍透過第三方代理與虛擬私人網路(VPN)持續存取。此外,Anthropic 先前指控中國實體,特別是阿里巴巴,從事「蒸餾」——即透過反向工程 AI 模型以複製其能力的過程。針對這些發現,Anthropic 工程師 Thariq Shihipar 表示,引起爭議的程式碼是 3 月開始的一項實驗性措施,旨在減輕未經授權轉售商對帳號的濫用,並防止上述的蒸餾行為。Shihipar 指出,該功能原定在隨後的版本中移除。
Institutional responses within China have been immediate. Reports indicate that Alibaba instructed its personnel to cease the professional utilization of Claude Code effective July 10, citing security imperatives.
中國國內機構的反應迅速。報告指出,阿里巴巴已指示其員工自 7 月 10 日起停止將 Claude Code 用於專業用途,理由是出於安全必要性。
Conclusion
Chinese regulators have urged the immediate remediation of Claude Code installations following the discovery of unauthorized data transmission capabilities.
中國監管機構在發現未經授權的數據傳輸功能後,促請立即修復安裝了 Claude Code 的系統。
Vocabulary Learning
The Architecture of 'Institutional Neutrality' & Nominalization
To transcend B2/C1 fluency and enter the C2 stratum, a learner must master the depersonalization of agency. The provided text is a masterclass in Nominalization—the process of turning verbs (actions) into nouns (concepts). This shifts the focus from who is doing something to what is occurring, creating an aura of objective, institutional authority.
◈ The Linguistic Pivot: From Action to State
Observe how the text avoids simple active constructions (e.g., "The government warned about...") in favor of complex nominal clusters:
- "Identification of Security Vulnerabilities" Instead of "Identifying vulnerabilities."
- "Systemic friction regarding intellectual property" Instead of "Systems are clashing over IP."
- "Immediate remediation of Claude Code installations" Instead of "Fixing the installations immediately."
C2 Insight: By utilizing nouns like remediation, identification, and friction, the writer removes the emotional volatility of the conflict. This is the hallmark of high-level diplomatic and technical discourse.
◈ Lexical Precision: The 'Hedge' and the 'Hammer'
C2 mastery requires the ability to calibrate the intensity of a claim. Note the interplay between precise technical jargon and calculated descriptors:
"...an experimental measure initiated in March to mitigate account abuse..."
- Mitigate (C2): Far more precise than reduce or stop. It implies making a problem less severe without necessarily eliminating it.
- Slated for removal (C2 Idiomatic): A professional collocation meaning 'scheduled.' It suggests an organized, bureaucratic timeline rather than a random decision.
◈ Syntactic Sophistication: The Appositive Expansion
Look at the phrase: "...the National Vulnerability Database (NVDB), an entity affiliated with the Ministry of Industry and Information Technology..."
This structure—[Noun] + [Appositive Phrase]—allows the author to embed critical context without breaking the sentence's momentum. A B2 student would likely use a separate sentence ("The NVDB is an entity..."). A C2 writer integrates the identity into the flow, maintaining a high density of information per clause.