US Cries Thief, China Retaliates with Total Block
The technology war between the United States (US) and China is heating up again. Recently, US-based Anthropic accused China’s Alibaba of illegally ‘stealing’ the capabilities of its Claude AI model.
Alibaba did not accept the accusation and immediately ordered a total block. All Alibaba employees are now prohibited from using Anthropic’s Claude Code at work.
Claude Code has drawn attention because its features can help identify users linked to China, according to a person familiar with the directive. The coding assistant tool is designed for software development.
The tool has become very popular among programmers in China, even though Anthropic has imposed access restrictions on users and entities in the country.
An internal source said Alibaba employees have been asked to switch to using the company’s own coding platform, named Qoder, Reuters reported on Monday.
Alibaba and Anthropic did not immediately respond to requests for comment from Reuters. Alibaba has not publicly responded to the accusations previously levelled by Anthropic.
Last month, Anthropic said it had suffered a ‘distillation’ attack from Alibaba. This attack involves training a less sophisticated model using the outputs of a more advanced model.
In a letter to two US senators seen by Reuters, Anthropic said the distillation attack helped accelerate the ability of Chinese AI models to match the capabilities of Anthropic’s own Mythos Preview.
Alibaba’s block on Claude Code emerged just days after developers said Claude Code had a mechanism that checks a user’s environment, including time zone and proxy-related information, and inserts subtle markers into prompts sent to Anthropic’s servers.
An Anthropic employee wrote on platform X that the feature was part of an experiment launched in March 2026. Its goal is to prevent account misuse by unauthorised resellers and protect against model distillation.
The source who spoke to Reuters about the Alibaba ban said Anthropic’s China-targeted restrictions are difficult to enforce against individual users, as they can use servers in the US and make data traffic appear as if it originates from there.
However, companies are more aware of the legal and compliance risks, the person added.
Due to the protective policies of US AI model developers seeking to prevent unauthorised access, resale, and distillation of their systems, Chinese cloud and AI companies have turned to domestic and open-source models such as DeepSeek, Alibaba’s Qwen, Moonshot, and Zhipu.
At the same time, Chinese AI models are increasingly penetrating the US market—a development that is raising concerns among some US industry experts.
Alibaba’s ban was first reported by Chinese media.