Alibaba Has Announced a Company Wide Ban on Claude Code

English | Chinese Original

Date: 2026-07-04

Yesterday Alibaba announced a complete ban on Claude Code.

Since July 10, internal employees have been completely prohibited from using this tool in the office environment, and the self-developed Qoder has been recommended as an alternative.

I think everyone should know the reason for this ban. It is because Claude Code was exposed to the security risk of implanting backdoors and accurately locked 147 Chinese domain names. Mainstream domestic technology companies and AI laboratories such as Meituan, NetEase, Baidu, Alibaba, and ByteDance are all on the monitoring list.

A complete review of several articles I wrote this week.

Claude Code poisoned the code in order to block Chinese users.

The ban of Club A’s account aroused public outrage, and even the big V couldn’t bear to start complaining.

I think Ali’s approach is very courageous.

This time Alibaba dared to completely ban Claude Code, which also gave me some attention and expectations for Qoder and Qwen large models, and I also hope to see Alibaba make a breakthrough in the field of AI.

Alibaba is very generous to its employees in terms of reimbursement amounts and the use of free model quotas, but of course this is also a major advantage of being a leading manufacturer and Internet company.

I have seen people from Internet companies post screenshots of token consumption. If they connect to the API, some people exaggerate that it can consume tens of thousands of dollars a week. This is still Opus 4.6. If it is converted to Fable 5, I dare not think about it.

In fact, if Company A targets us like this every day, it will force foreigners to have the courage to overtake in a corner.

The last time this was done to us, it was a technological blockade and we had no choice but to engage in innovation. At that time, the chip was jammed and the operating system was jammed, so we were forced to find a way to localize it.

Today, AI is at it again. At the software level, it will bury backdoors, tag you, and ban your accounts in a discriminatory manner.

History is always surprisingly similar.

The feeling of being stuck in the neck is unpleasant, but every time we are stuck, it forces us to make progress.

In fact, the Chinese model is not bad anymore. In the past two days, I saw an evaluation of GLM-5.2 and Fable 5. Someone used these two models to compare the simulated 3D earth. It can be seen that the gap in model capabilities has been very small.

But it’s like what DeepSeek said before when it released V4: Don’t be tempted by praise, don’t be afraid of slander, do what is right, and be honest.

This is true for chips, operating systems, and large AI models.

I hope this time I can force the national models to improve, not just Alibaba.

Let everyone absorb cheap and easy-to-use tokens.