China alleges that Claude Code contains backdoors, calls mechanism 'a serious threat' — Gov't claims Claude sends sensitive information to remote servers without consent

Summary

China's NVDB flagged Claude Code for security vulnerabilities, alleging it sends sensitive data without consent. Anthropic confirmed an experiment to prevent abuse, which is being rolled back. Despite restrictions, Chinese developers access the tool, prompting the update advisory.

Key Takeaways

1. China's National Vulnerability Database (NVDB) stated that Claude Code contains 'security backdoor vulnerabilities' and advised users to uninstall or update the tool.
2. Versions of the tool released between April and June 2026 can send sensitive information such as user location and identity to remote servers without the user’s consent due to a built-in monitoring mechanism.
3. Anthropic engineer Thariq Shihipar confirmed on X that it was an experiment the company launched in June 'to prevent account abuse from unauthorized resellers and protect against distillation.'

Statement Breakdown

• Claimed Facts: 60% of statements the article presents as facts
• Opinions: 30% of statements classified as editorial or subjective
• Claims: 10% of statements surfaced for additional reader evaluation

Credibility & Bias Reasoning

Credibility assessment: The article presents information from official Chinese government statements and Anthropic's own engineer, providing a balanced view. However, it relies on a single news source and lacks direct quotes from Anthropic's official statements, limiting its overall credibility.

Bias assessment: Geopolitical Tech Scrutiny. The article frames the situation through the lens of China's national security concerns and allegations against a US-based AI company. It highlights the geopolitical tension surrounding AI development and data security.

Note: This article reports on allegations and official statements from China regarding AI security. It's important to note that the AI tool is not approved for public use in China, and Anthropic has restricted its use in the region.

Credibility flag: Information Verified

Claimed Facts (5)

• This is a direct report of an official statement from a government database.
• This presents information attributed to a reputable news source regarding specific technical details.
• This is a factual report of a statement made by an Anthropic employee on a public platform.
• This is a factual report of a specific commitment made by an Anthropic engineer.
• This states a factual action taken by Anthropic regarding its services in China.

Opinions (5)

• The phrase 'It seems that' indicates an interpretation rather than a confirmed fact.
• The phrase 'It seems that' indicates an interpretation of the situation.
• The word 'probably' indicates speculation about the reasoning behind the agency's recommendation.
• The word 'curious' expresses a subjective judgment about the guidance.
• The word 'seemingly' indicates an interpretation of Beijing's actions.

Claims (5)

• This is a headline that uses strong, accusatory language ('alleges', 'serious threat') which can be sensationalized.
• This is a headline that uses strong, accusatory language ('claims') which can be sensationalized.
• This statement highlights a potential contradiction or unusual circumstance without further explanation, which could be misleading.
• The claim of '24,000 fraudulent accounts' is a specific, large number that lacks immediate substantiation within the text and could be an exaggeration or unverified accusation.
• This is a repeated accusation of 'distillation' without detailed evidence provided in the article, relying on Anthropic's prior claims.

Key Sources

• China's National Vulnerability Database (NVDB) — Government Database
• Wall Street Journal — News Publication
• Thariq Shihipar — Anthropic engineer
• Anthropic — AI Company
• Tom's Hardware — Tech News Website
• DeepSeek — AI Developer
• Alibaba — Technology Company