Alibaba restricts staff access to Anthropic's Claude Code

Anthropic has been aggressively closing loopholes that previously allowed Chinese companies to bypass restrictions on their AI models. Recent reports indicate that the company implemented covert tracking within Claude Code to identify users based in China. Thariq Shihipar of Anthropic described this mechanism as a temporary experiment launched in March to curb unauthorized reselling and prevent model distillation, where developers train their own AI systems using the outputs of established models.

Alibaba has responded by labeling the Claude tool as high-risk software. The company is now mandating that staff transition to Qoder, its proprietary coding assistant, to maintain development security. This shift highlights the growing tension between American AI restrictions and the internal software requirements of Chinese tech conglomerates.