This is the clearest sign yet that the US-China AI relationship is fracturing at the enterprise level, not just the export-control level. A Western coding tool just got blacklisted inside one of China's largest tech companies.
- Alibaba will ban employees from using Anthropic’s Claude Code starting July 10, classifying it as high-risk software, CNBC reported.
- Staff must uninstall Anthropic’s models and agent products, switching instead to Alibaba’s own Qoder assistant.
- Anthropic told the US Senate in June that Alibaba conducted the largest known distillation attack against it to date.
- Separately, developers found Claude Code contained code that could detect if users were based in China, which Anthropic says it built to fight account abuse and distillation.
- Anthropic’s Thariq Shihipar said the team had been meaning to remove that detection code “for a while.”
What Happened
Alibaba will bar employees from using Anthropic’s Claude Code starting July 10, classifying it as high-risk software, TechCrunch reported. Staff must uninstall Anthropic’s models and agent products and switch to Alibaba’s own Qoder assistant.
The ban follows a June letter Anthropic sent to the US Senate Committee on Banking, Housing, and Urban Affairs. Anthropic accused Alibaba of conducting the largest known distillation attack against it to date, a practice where a model’s outputs train a rival system at lower cost.
Separately, developers flagged that a version of Claude Code contained hidden code able to detect whether users were based in China. Anthropic’s Thariq Shihipar said on X this was an experiment launched in March to fight account abuse and protect against distillation.
Anthropic already prohibits Chinese companies from using its models. The Financial Times separately reported Anthropic is closing loopholes that let Chinese firms access Claude through third-country entities, including one case involving Ant Group’s Singapore-based operations.
Why It Matters
This is a concrete instance of the US-China AI relationship splitting at the corporate level, not just through government export controls. Alibaba losing access to a leading coding tool, and Anthropic losing a foothold inside one of China’s largest tech companies, is a clean preview of what a fully bifurcated AI market looks like in practice.
Neither side has published hard evidence. Alibaba has not detailed its distillation findings publicly, and Anthropic has not released technical proof beyond its Senate letter. The hidden China-detection code in Claude Code, whatever its original intent, gives Alibaba a legitimate-sounding security justification independent of whether the distillation claim itself holds up. Anthropic’s own precedent here matters too, it made similar distillation accusations against DeepSeek, MiniMax, and Moonshot AI in February.
“We’ve actually been meaning to take this down for a while.”
Thariq Shihipar, Anthropic
Bottom Line
Watch whether other major Chinese tech firms, Tencent or ByteDance among them, follow Alibaba’s lead. A second or third ban would confirm this is becoming a sector-wide pattern rather than one company’s isolated security decision.
For SaaS teams with any operations touching both US and Chinese markets, this is a direct compliance signal. Employees in different regions may soon need genuinely different AI tooling, not just different account permissions.
