Claude Code has become the center of a two-way restriction fight involving Anthropic and major Chinese technology companies. On one side, Anthropic is trying to block Chinese-controlled companies from using the tool. On the other, Alibaba is reportedly telling its own employees to stay away from Claude Code and remove Claude models.
The result is a complicated access problem. Anthropic’s rules are meant to limit who can use Claude Code, but companies are reportedly finding routes around those limits. Meanwhile, concerns inside China appear to be producing bans of their own.
What Anthropic is trying to block
According to the Financial Times, Anthropic is trying to prevent Chinese companies from accessing Claude Code. The company’s terms of service explicitly ban sales to companies controlled by China.
That restriction is clear in principle, but the source article describes several ways companies are reportedly getting around it. Companies like Ant Financial and ByteDance are said to be using cloud services, overseas subsidiaries in Singapore, or VPNs to bypass the limits.
This creates a practical enforcement challenge. A formal ban on sales can define which customers are not allowed, but the source shows that access can still happen through indirect routes. When a company can connect through another service, a foreign subsidiary, or a VPN, the question becomes not only what the rules say, but how reliably they can be enforced.
Alibaba’s ban moves in the opposite direction
The pressure is not coming only from Anthropic. The Information reports that Alibaba is banning its own employees from using Claude Code and requiring them to delete all Claude models.
That makes the Claude Code China issue unusual. Anthropic is trying to keep some Chinese-controlled companies out, while at least one major Chinese company is reportedly moving to keep Claude Code out of its own internal environment.
The ban follows reports of hidden code in Claude Code that could flag users based in China or linked to a Chinese lab. The source article does not say that Alibaba’s decision was based on any additional technical finding beyond those reports. It does, however, connect the internal ban to concerns about that hidden code.
The hidden-code explanation
Anthropic’s Thariq Shihipar called the hidden-code mechanism "an experiment from March". According to the source article, he said it was intended to stop account abuse and distillation.
He also said stronger safeguards have since replaced it. That detail matters because it frames the reported code not as the current main protection system, but as an earlier experiment that Anthropic says has been superseded.
Still, the reported existence of code that could flag users based in China or linked to a Chinese lab helps explain why Chinese companies may view Claude Code as sensitive. Even if Anthropic describes the mechanism as part of an anti-abuse effort, the same fact can look different from the perspective of a company deciding what tools its employees may use.
Why distillation sits at the center
The source article identifies distillation as one of the central concerns behind Anthropic’s actions. Distillation, as described there, means training smaller models on Claude’s outputs.
Anthropic has previously accused Alibaba, DeepSeek, Moonshot AI, and MiniMax of using Claude for distillation. Those accusations are part of the backdrop to the current restrictions and safeguards around Claude Code.
Seen through that lens, Anthropic’s position is about protecting access to Claude outputs and preventing account abuse. The company is trying to stop restricted users from reaching Claude Code, while also trying to prevent its models from being used to train competing systems.
For the companies named in the source article, the issue is not only whether Claude Code is technically useful. It is also whether using it creates exposure to restrictions, monitoring, or future enforcement. That helps explain why a company could be both a target of access restrictions and a company that restricts the tool internally.
A messy enforcement problem for AI tools
The facts in the source article point to a broader tension around advanced AI tools. Terms of service can restrict sales to companies controlled by China, but users may still attempt access through cloud services, overseas subsidiaries in Singapore, or VPNs.
At the same time, companies that might benefit from Claude Code may decide that the risks are too high for internal use. Alibaba’s reported order to delete all Claude models shows how quickly an access fight can become an internal compliance issue.
Claude Code’s China problem is therefore not a simple story of one company banning another. It is a layered dispute over access, safeguards, distillation, and trust. Anthropic is trying to control who can use its coding tool, while Chinese companies are responding both by working around restrictions and, in Alibaba’s case, reportedly banning the tool from inside.