Alibaba Restricts Staff Access to Anthropic’s Claude Code Amid Rising AI Geopolitics

In a significant shift for its internal development operations, Alibaba has officially moved to block its workforce from utilizing Anthropic’s AI-powered programming assistant, Claude Code. Effective July 10, the tech giant has categorized the tool as a high-risk application, mandating that its engineers pivot toward internal alternatives, specifically the company’s proprietary “Qoder” platform.

The Growing Divide in AI Accessibility

This directive arrives against a backdrop of tightening restrictions regarding the use of Western-developed artificial intelligence models within China. Anthropic, the developer behind the Claude series, has maintained a strict policy prohibiting Chinese corporations-and their international subsidiaries-from accessing its technology. As global AI competition intensifies, the company has been aggressively patching vulnerabilities that previously allowed users in restricted regions to bypass these geographic barriers.

The geopolitical landscape of AI is becoming increasingly fragmented. With the U.S. government continuing to tighten export controls on high-end semiconductors and advanced AI software, companies like Anthropic are under mounting pressure to ensure their intellectual property does not inadvertently support the technological ambitions of restricted nations. This trend mirrors the broader “decoupling” seen in the semiconductor industry, where supply chains are being bifurcated to align with national security interests.

Uncovering the “Secret” Detection Mechanism

The tension surrounding these access restrictions recently surfaced on social media platforms like Reddit, where users discovered that certain iterations of Claude Code appeared to be running background processes designed to identify and flag Chinese-based users.

Thariq Shihipar, a representative for Anthropic, addressed the controversy on X (formerly Twitter), clarifying that the code was part of a limited-scope experiment initiated in March. According to Shihipar, the primary objective was to curb account abuse by unauthorized third-party resellers and to prevent “model distillation”-a process where developers attempt to train smaller, cheaper AI models using the high-quality outputs generated by more sophisticated systems like Claude.

“The team has implemented more robust security measures since that time, and we have been planning to sunset that specific detection feature for a while,” Shihipar noted, emphasizing that the company’s focus remains on protecting its proprietary infrastructure rather than targeted surveillance.

Alibaba’s Strategic Pivot to Qoder

For Alibaba, the decision to ban Claude Code is as much about corporate security as it is about promoting its own ecosystem. By labeling the tool as “high-risk,” the company is effectively insulating its internal codebase from potential external oversight or data leakage.

This move reinforces Alibaba’s commitment to its own AI stack. By funneling its engineering talent toward Qoder, the company is not only mitigating compliance risks but also accelerating the adoption of its domestic AI tools. As China continues to invest heavily in its own large language models (LLMs) to compete with Silicon Valley, the reliance on homegrown solutions like Qoder is expected to become the standard for major Chinese tech firms, further cementing the divide between Western and Eastern AI development environments.