China's MIIT Warns Claude Code Has Security Backdoor; Alibaba Has Already Banned Its Use
Miles Bennett
China's official vulnerability database says Anthropic's AI coding tool Claude Code carries a built-in surveillance backdoor that can transmit user location and identity data without consent; Alibaba has already banned staff from using it, putting dev-tool supply-chain security squarely on regulators' radar.
What exactly did China's MIIT find?
On July 8, China's National Vulnerability Database (NVDB) — run by the Ministry of Industry and Information Technology — issued an alert: Claude Code contains a built-in monitoring mechanism that can send users' geographic location and identity markers to a remote server without their consent.
Affected versions span 2.1.91 through 2.1.196, a broad range.
This means → The agency is not calling this a gray-area data-collection practice. It is classifying it as a "backdoor" — a significantly more serious designation than vague privacy-policy concerns.
What is the NVDB telling users to do?
Immediately audit every development terminal running the affected versions, and uninstall or upgrade to the latest version with the backdoor code removed.
Tighten outbound-connection permissions and traffic monitoring for dev tools on core business network segments to prevent unauthorized data exfiltration.
In plain terms = strip out the suspect versions first, then lock down the channels through which dev tools can phone home.
Why had Alibaba already moved first?
Reuters previously reported that Alibaba banned employees from using Claude Code at work because the tool has features that can help identify Chinese users.
Anthropic did not immediately respond to Reuters' request for comment.
This means → Alibaba's ban preceded the MIIT alert, suggesting the corporate sector sensed the risk before the regulator acted publicly.
What does this mean for developers and the wider industry?
The alert pushes dev-tool supply-chain security to the front of the regulatory agenda — previously, scrutiny focused on chips and operating systems; now AI coding tools are in the crosshairs too.
Whether Anthropic can rebuild trust in the Chinese market depends on its formal response to the security allegations.
This reflects a fast-rising compliance bar for foreign AI tools in China amid the broader U.S.–China tech rivalry — the question is no longer "can we use it" but "do we dare."
Content is for reference only, not financial advice.