Alibaba Just Banned an AI Coding Tool Built by the Same Company That Already Banned Alibaba's Country.

 

Starting July 10, Alibaba employees will reportedly be banned from using Claude Code, Anthropic's programming tool, according to multiple reports. The stated reason: Alibaba has classified it as high-risk software and is pushing staff toward its own internal tool, Qoder, instead.

Here's what makes this genuinely interesting rather than just another corporate IT policy update: the ban runs in both directions.

Anthropic already prohibits Chinese companies, and foreign entities owned by them, from using its models at all. That's a standing policy, not a new one. What's new is the reported reason feeding into Alibaba's decision. A Reddit post surfaced claims that a version of Claude Code was quietly able to identify Chinese users behind the scenes, as part of Anthropic's effort to close loopholes that let restricted users access Claude anyway.

Anthropic's response, from staff member Thariq Shihipar, doesn't deny this happened. He confirmed it was an experiment launched in March, intended to prevent account abuse from unauthorized resellers and to protect against distillation, the practice of training a separate AI model on another model's outputs. He also said stronger mitigations have since replaced it, and that the identifying mechanism was already scheduled for removal.

So neither side is really disputing the core facts. Anthropic built something to detect unauthorized access from a market it already restricts. Alibaba is responding by banning the tool entirely and framing it as a security risk to its own employees.

Whether you read this as "AI companies are right to police unauthorized use of their models" or "this is exactly the kind of quiet backend behavior enterprises should be worried about" probably depends on which side of the access restriction you're standing on.

Either way, it's a clean example of a bigger, messier problem the AI industry hasn't solved yet: how do you enforce national-level usage restrictions on a product that's fundamentally software, without building something that looks a lot like surveillance to the people on the other end of it.

Does this change how you think about enterprise AI tools built by companies with their own geopolitical restrictions baked in?

#AI #TechPolicy #Anthropic #Alibaba #AIGovernance