China pushes local coding tools after security alert against Claude Code

What happened?

China's National Vulnerability Database, overseen by the Ministry of Industry and Information Technology, issued an alert this week stating that multiple versions of Claude Code contain a 'backdoor' that could compromise national security. The alert did not provide specific technical details, but according to analysts cited by the South China Morning Post, it has already prompted Chinese developers to begin migrating toward domestic coding tools. Claude Code, launched by Anthropic in 2025 as a direct competitor to GitHub Copilot, is a coding assistant based on the Claude model that allows developers to generate, debug, and refactor code using natural language. The Chinese warning adds to growing regulatory scrutiny of foreign AI tools in the country.

Why it matters

This action is not an isolated incident. It is part of China's ongoing effort to reduce dependence on foreign technologies, especially in critical areas like artificial intelligence. The alert against Claude Code could accelerate the adoption of local alternatives such as CodeGeeX from Zhipu AI, Fitten Code from ByteDance, or Tongyi Lingma from Alibaba. According to data from consulting firm IDC, China's AI coding assistant market reached $1.2 billion in 2025, with annual growth of 45%. Local tools like CodeGeeX already have over 2 million registered users, according to Zhipu AI. For global tech companies, this represents further fragmentation of the development tools market, with implications for standards, interoperability, and data sovereignty. Companies like Microsoft (with GitHub Copilot) and Google (with Gemini Code Assist) could also be affected if China extends similar restrictions.

Historical context and comparisons

The situation echoes the ban on Google in 2010 and the blocking of services like GitHub during certain periods. However, the current context is different: generative AI has become a geopolitical battleground. The alert against Claude Code parallels U.S. restrictions on AI chips to China and reinforces the narrative that AI technology is not just a commercial issue but one of national security. In 2023, China had already placed Anthropic on a list of AI companies subject to security reviews, according to leaked documents. Additionally, in 2024, the Chinese government issued guidelines requiring AI models used in the country to be trained on local data and comply with socialist values. The current alert is another step in this direction, similar to how Russia promoted its own coding assistant, GigaCode, as an alternative to Western tools after sanctions in 2022.

Consequences for companies and developers

For Chinese developers, transitioning to local tools may involve a learning curve and potential short-term productivity losses. However, in the long term, it could foster a more autonomous AI ecosystem. According to a survey by software analytics firm JetBrains, 68% of Chinese developers used foreign coding assistants in 2025, primarily GitHub Copilot and Claude Code. After the alert, this figure is expected to drop to 30% by 2027. For global companies operating in China, the alert introduces regulatory uncertainty and could force a reassessment of Claude Code use in their local teams. For example, companies like BMW and Apple, which have R&D centers in China, may have to migrate their development workflows. Additionally, the move could inspire similar actions in other countries seeking to protect their domestic tech industries, such as India, which has already shown interest in promoting its own coding assistant, Bhashini, for regional languages.

What readers should know

• The alert does not provide concrete evidence of the 'backdoor,' raising doubts about its technical basis versus political motivations. Cybersecurity experts like researcher Xiao Zhang have noted there is no public evidence of vulnerabilities in Claude Code that would justify a national security alert.
• Tools like Claude Code are not officially blocked, but regulatory pressure may lead to voluntary abandonment by Chinese companies to avoid potential sanctions.
• Local alternatives like CodeGeeX have shown significant progress but still do not match Claude Code's maturity in complex tasks. In internal benchmarks by Zhipu AI, CodeGeeX achieved 78% accuracy in code generation, compared to Claude Code's 85%.
• The global AI coding assistant market is valued in the billions, and fragmentation could slow innovation. According to Gartner, the market will reach $8 billion by 2028, but regulatory fragmentation could reduce growth by 15%.

Market impact and projections

The market share of foreign coding tools in China is expected to decline sharply, while local tools could capture over 80% of the market within the next two years, according to estimates from analysts at Counterpoint Research. This could incentivize Anthropic and other companies to adapt their products to comply with Chinese regulations, although geopolitical tensions make this uncertain. For example, Anthropic could offer a version of Claude Code hosted on Chinese servers and trained on local data, similar to what Microsoft did with Azure in 2014. However, any concession could be seen as a sign of weakness in the West. Globally, market fragmentation could lead to divergent standards, as already seen with language models (e.g., China's responsible AI standard vs. the EU framework).

Conclusion

The alert against Claude Code is a milestone in the growing technological divergence between China and the West. For industry professionals, it is crucial to monitor these developments, as they will redefine the available tools and software development strategies globally. China's decision not only affects Anthropic but sends a signal to the entire AI ecosystem: technological self-sufficiency is a strategic priority. Meanwhile, developers will need to prepare for a multi-tool environment, where the choice of coding assistant increasingly depends on geography and regulation rather than technical capabilities.