China's coding tools are experiencing growth as Claude Code faces scrutiny.

      In summary, China's National Vulnerability Database has identified several versions of Anthropic's Claude Code as having a security "back door" that may transmit user locations and identifiers to external servers without consent. Anthropic stated that the code was part of an experiment aimed at preventing unauthorized model distillation and that its policies have already prohibited users in China from accessing its services. Analysts suggest that this controversy will accelerate the transition of Chinese developers to local coding tools such as ByteDance’s Trae, Alibaba’s Qoder, Tencent’s CodeBuddy, and Zhipu’s CodeGeeX and ZCode.

      Beijing's cybersecurity warning regarding Anthropic is anticipated to hasten this ongoing shift. Analysts informed the South China Morning Post that Chinese developers are increasingly turning to domestic coding tools.

      The National Vulnerability Database, managed by the Ministry of Industry and Information Technology, issued a warning this week alleging that certain versions of Claude Code contain a security "back door." The agency indicated that the software could send user locations and identity information to remote servers without user consent, advising local organizations to either uninstall the affected versions or upgrade to secure updates.

      Regarding the situation, both parties acknowledge the presence of the code but differ in its intended purpose. Anthropic admitted to including a tracking feature in Claude Code, characterizing it as a measure against unauthorized model copying. The company reiterated that its usage policy has always excluded users based in China.

      The dispute has roots in prior conflicts, as last month, Anthropic accused Alibaba of attempting to extract its AI capabilities, and subsequently, Alibaba has prohibited its staff from using Claude Code, labeling it as high-risk software.

      Cai Peng, a cybersecurity partner based in Beijing at Zhong Lun Law Firm, anticipates that more Chinese companies will abandon foreign AI tools due to increasing security concerns and the country’s imperative for technological self-reliance.

      ByteDance’s Trae functions as an AI-native development environment centered on natural-language prompts and automated agents. Similar to Claude Code, it generates code, debugs software, clarifies codebases, and can create applications from scratch. ByteDance claims that by the end of 2025, Trae had over 6 million registered users and 1.6 million monthly active users. Trae emphasizes localization, supporting domestic foundation models such as Doubao and DeepSeek, and permitting users to integrate their own model APIs.

      Alibaba’s Qoder, which evolved from its Tongyi Lingma assistant, is designed as a comprehensive development workspace instead of merely an assistant. It manages code generation, project analysis, and autonomous task execution, allowing users to incorporate their own models through customized API keys.

      Tencent and Zhipu are also significant players in this space. Tencent Cloud’s CodeBuddy operates on the company’s Hunyuan model and DeepSeek’s, supporting over 200 programming languages while providing functionalities such as code completion, debugging, refactoring, multi-file generation, and testing. Zhipu AI features two tools: CodeGeeX, one of China’s earliest coding assistants trained using Huawei’s MindSpore framework on Ascend processors, affirming hardware independence; and ZCode, a more ambitious agentic workspace designated as the official environment for Zhipu’s GLM-5.2 model, boasting a context window of up to 1 million tokens and capable of orchestrating workflows from requirements to documentation. Notably, ZCode still provides a multi-model interface that allows for switching to OpenAI and Google models during projects, indicating that complete separation is not yet achieved.

      Ultimately, the controversy regarding the back door reflects a broader trend of a software ecosystem dividing along national lines. The Chinese alternatives are becoming increasingly competitive, with affordable domestic models narrowing the gap with U.S. frontier labs. This development diminishes the primary reason for Chinese developers to endure the complications of using American tools. The pressure is reciprocal, as Beijing contemplates restrictions on foreign access to its best models while Washington limits what U.S. labs can sell.

      Claude Code has never been a particularly suitable option for cost-sensitive buyers, as illustrated by Microsoft’s discreet withdrawal from it. Factors such as pricing, not just political tensions, are steering developers toward more affordable alternatives. As a consequence, two distinct ecosystems are emerging, increasingly separate from one another. For Anthropic, a market it never officially catered to is now off-limits, while domestic tools are becoming more readily available.