China's government has issued a security warning on Anthropic's AI coding tool Claude Code, increasing the likelihood that Chinese developers will switch faster to domestic alternatives. With security concerns intersecting with a push for technological self-reliance, China's AI coding tools market could be reshaped from a foreign-tool focus to locally based platforms.

On July 11 local time, the South China Morning Post reported that China's National Vulnerability Database (NVDB) warned this week that several versions of Claude Code include a security "backdoor".

NVDB claimed the software could transmit location and identity information to remote servers without user consent. It advised institutions and companies in China to immediately delete affected versions or apply security patches. The warning came after Anthropic recently acknowledged inserting code into Claude Code to track illegal distillation of its models, or unauthorised copying.

On July 9, Anthropic said in response to the NVDB warning that access to its service from users in China had long been prohibited under its usage policy. The company said it has maintained a policy of not allowing China-based users and that the move did not newly restrict users in China.

In China, some expect the incident to accelerate the departure from foreign AI developer tools. Partner Sai Feng (赛锋) of Beijing Zhonglun Law Firm analysed that as security concerns grow and overlap with China's strategic need for technological self-reliance, more companies may reduce their use of overseas AI tools.

The first service cited as an alternative is ByteDance's Trae. Trae is an integrated development environment that uses natural-language commands and autonomous AI agents. It supports code generation and debugging, explanations of complex code and initial application setup.

ByteDance said Trae had 6,000,000 registered users and 1,600,000 monthly active users as of the end of 2025. The China version supports locally based models such as Doubao and DeepSeek, and is designed to allow users to connect their own model APIs. The structure could align with demand from Chinese companies seeking to reduce reliance on foreign models.

Alibaba's Coder CN is also cited as a key alternative. Coder CN is an agent-based coding platform developed from the existing AI coding tool Tongyi Lingma and is designed more like an AI-based development workspace than a simple code assistant. It supports code generation, project analysis and autonomous task execution, and overseas users can connect the base model they want through a customised API key. It can also integrate with external enterprise development tools.

Tencent Cloud's CodeBuddy uses Tencent's Hunyuan large language model together with the DeepSeek model. It supports more than 200 programming languages including Python, Java, Go and JavaScript, and provides context-based code completion, debugging and refactoring functions. It is also designed to break down software requirements and support code generation across multiple files and testing work.

ZhipuAI's coding tools are also drawing attention. CodeGeeX is one of China's early AI coding tools and has built a local developer community based on compatibility with major integrated development environments. ZhipuAI has stressed that CodeGeeX is an alternative based on domestic technology, saying it was trained on Huawei's MindSpore framework and Ascend AI processors. Another tool in the same lineup, ZCode, is a workspace that automates the entire software development process from requirements planning to debugging, testing and documentation.

It was built as the official development environment for ZhipuAI's GLM-5.2 model and, it said, supports a context window of up to 1,000,000 tokens to improve the efficiency of analysing large codebases. It also provides a multi-model interface that lets users switch within a single project between Chinese models and models from OpenAI and Google.

The industry views the warning as something that could spur structural changes in China's AI development ecosystem, beyond a simple security dispute. In particular, if Chinese companies begin to place greater weight not only on coding productivity but also on control over data, integration with local infrastructure and regulatory compliance, the market share of domestically made AI development tools could rise quickly.

Key points to watch are whether the warning will lead to actual tool replacement by companies, and how quickly local companies such as ByteDance, Alibaba, Tencent and ZhipuAI can absorb the gap left by Claude Code.