小互 (X)
2 sources愈演愈烈了 中国工信部发布关于防范Claude Code安全后门隐患的风险提示 公告称：Claude Code2.1.91至2.1.196版本内置了监控机制，未…
Original
中国工信部发布风险提示，指出Claude Code 2.1.91至2.1.196版本存在安全后门，内置监控机制，未经用户同意回传敏感信息。建议立即卸载或升级，并加强权限管控与流量监测。
Back to feed
News Story
SynthePulse Insight · AI deep reading
Claude Code 'Backdoor' Controversy: China's MIIT Warning vs. Anthropic's Rebuttal
Version 1 · 2 sources
China's Ministry of Industry and Information Technology (MIIT) National Vulnerability Database issued a risk alert, claiming that certain versions of Claude Code contain built-in monitoring mechanisms that can transmit sensitive information such as user location and identity identifiers without user consent. Anthropic responded that the feature was an anti-abuse protection experiment and emphasized that Claude is not authorized for use in China. The incident reflects the complex interplay of security, compliance, and geopolitics between US and Chinese AI tools.
- China's MIIT National Vulnerability Database announced that Claude Code versions 2.1.91 to 2.1.196 contain built-in monitoring mechanisms that can transmit sensitive information such as user location and identity identifiers to remote servers.
- Anthropic rebuts that the feature was an anti-abuse protection experiment conducted earlier this year to prevent model distillation, not a backdoor.
- Anthropic reiterates that Claude is not authorized for use in China, and its policy prohibits use by entities controlled from China.
- China's warning recommends users uninstall affected versions or upgrade to the latest version with the backdoor code removed, and strengthen external connection controls for development tools.
- The incident follows Anthropic's accusation last month that Alibaba attempted to extract its AI capabilities; Alibaba has since asked employees to stop using Anthropic tools starting July 10.
- For multinational companies, regional compliance review of AI coding tools has become a governance challenge, and a single global policy may be insufficient.
Open section navigationChina's Official Warning: Security Backdoor in Claude Code
China's Official Warning: Security Backdoor in Claude Code
On July 8, 2026, China's MIIT National Vulnerability Database issued a risk alert stating that Anthropic's AI coding assistant Claude Code versions 2.1.91 to 2.1.196 contain built-in monitoring mechanisms that can transmit sensitive information such as user location and identity identifiers to remote servers without user consent. The database recommends that relevant units and users immediately uninstall affected versions or upgrade to the latest secure version with the backdoor code removed, while strengthening external connection permission controls and traffic monitoring for development tools in core business network segments.
The warning was issued by the MIIT as an official security risk alert but did not provide specific technical details or independent verification.
Anthropic's Response: Anti-Abuse Protection, Not a Backdoor
Anthropic disputes China's allegations. According to CNBC, Anthropic stated that the so-called 'backdoor' was actually an experiment conducted earlier this year to prevent model distillation—a process where the output of a large AI model is used to train another model. Anthropic emphasized that the feature was an anti-abuse protection measure, not a backdoor.
Additionally, Anthropic reiterated that Claude is not authorized for use in China, and its policy prohibits use by entities controlled from China. This stance directly opposes China's warning, with both sides providing entirely different characterizations of the same feature.
Geopolitical Context: Escalating US-China AI Disputes
This incident is not an isolated security issue but part of the broader US-China AI dispute. According to CNBC, Anthropic last month accused Alibaba of attempting to extract its AI capabilities. Alibaba did not comment at the time but has since asked employees to stop using Anthropic tools for work starting July 10, 2026.
The timing of China's warning closely follows Anthropic's accusation, pushing the issue beyond mere software telemetry into a broader geopolitical contest.
Practical Impact on Multinational Companies
For multinational companies with development teams in China or the Asia-Pacific region, the Claude Code warning highlights the regional compliance governance challenges of AI coding tools. AI coding assistants not only affect productivity but also directly access source code, internal systems, and company data. Security teams need to assess the types of data collected by the tool, data flow destinations, and whether regional rules permit employee use.
According to the South China Morning Post, security experts expect Chinese companies to treat AI vendors as strategic supply chain providers rather than mere software vendors. This means a single global AI tool policy may be insufficient; companies need to develop region-specific rules for different markets, including approved coding assistants, telemetry controls, network access, and vendor risk reviews.
Credibility boundary
This article is based on China's MIIT official announcement, Anthropic's response via CNBC, and reports from TechRepublic. The original Chinese warning announcement did not provide technical details or independent verification; Anthropic's response is a unilateral statement. The geopolitical background of the incident comes from CNBC reports, where Anthropic's accusation against Alibaba has not been independently confirmed.
Insight takeaway
The Claude Code 'backdoor' controversy is essentially a convergence of US-China AI technology competition and security review. China's official warning and Anthropic's rebuttal present conflicting claims without third-party technical verification. For enterprises, regardless of the feature's characterization, regional compliance risks are already real, and AI tool usage strategies must be adjusted according to local regulations.
Primary report
小互 (X)
Primary source
Same-event coverage
Also covered by 1 sources