Chinese tech giant labels Anthropic’s coding tool “high-risk,” pushes employees toward homegrown alternative Qoder
Alibaba is all set to stop its employees using Claude Code, a popular AI tool for coding that is developed by Anthropic, with effect from July 10 due to its internal classification of the tool as high-risk software. Following the announcement, the employees have been asked to shift to Qoder, Alibaba’s own AI coding assistant as a replacement.
Anthropic’s own restrictions — and a walk-back
The move comes even as Anthropic maintains its own ban on Chinese companies and foreign entities they own from accessing its models and tools. That policy has reportedly had gaps, allowing some China-based users to keep using Claude despite the restrictions.
To close those gaps, Anthropic is said to have quietly deployed a version of Claude carrying hidden tracking code designed to flag users’ locations and any ties to Chinese firms, per a report from The Information. The company has since pulled back from the effort.
Anthropic’s Thariq Shihipar addressed the matter on X, describing it as a March-era experiment aimed at curbing account abuse by unauthorized resellers and guarding against distillation, adding that stronger safeguards have since been put in place and the feature was already due to be retired.
A widening rivalry
The dispute unfolds against a backdrop of intensifying US-China competition in AI, with Chinese firms rolling out cheaper, competitive models. Z.ai’s GLM-5.2, released last month, has drawn attention in Silicon Valley for coding and agentic abilities that rival top US systems at a much lower cost — prompting comparisons to an earlier “DeepSeek moment.”
Washington’s export controls on advanced models, including Anthropic’s Claude Fable 5 and Mythos 5, have also fueled interest in Chinese alternatives, though Anthropic confirmed this month that the Commerce Department has since lifted those restrictions.
