Alibaba Implements Ban on Claude Code
A notice from Alibaba Group Holding has informed employees that they are no longer permitted to use Anthropic's Claude Code for company work. The Chinese technology conglomerate announced the ban, effective July 10, following an internal review that designated the AI coding tool as "high-risk software with security vulnerabilities."
Hidden Code Allegations Trigger Review
The decision stems from recent allegations that Anthropic, a prominent U.S. artificial intelligence company, embedded a tracking mechanism within its Claude Code. Security researchers initially flagged this practice on platforms like Reddit and GitHub, claiming the code was designed to identify users located in China or those affiliated with Chinese AI laboratories. According to cybersecurity publication International Cyber Digest, the alleged code collected signals such as Chinese time-zone settings and proxy server usage, transmitting this data back to Anthropic via concealed messages.
Anthropic Responds to Criticisms
Anthropic engineer Thariq Shihipar addressed the controversy on social media, characterizing the feature as an "experiment" introduced in March. Shihipar stated its purpose was to deter misuse by unauthorized resellers and prevent the distillation of Anthropic's advanced models—a process where outputs from a larger AI are used to train smaller, cheaper alternatives. Anthropic had previously accused Chinese AI firms, including Alibaba, of engaging in such distillation practices. Shihipar confirmed that the tracking mechanism would be removed as part of a planned re-release of Claude Fable 5, asserting that the company had intended to discontinue it after developing more robust safeguards.
Broader Context: US-China Tech Tensions
This incident unfolds amidst increasing restrictions on the deployment of advanced U.S. AI systems internationally. Last month, Washington imposed export controls on Claude Fable 5, the public iteration of Anthropic's more potent Mythos model, after Amazon researchers reportedly identified security weaknesses. Anthropic temporarily disabled the model before restoring access post-restrictions. Alibaba's move also aligns with a wider trend among Chinese tech companies to lessen their dependence on U.S. software and AI solutions. The company is reportedly recommending its own AI coding agent, Qoder, as an alternative. While Anthropic's products are not officially available in China, developers have found workarounds to access their coding and reasoning functionalities.