Alibaba Prohibits Use of Anthropic’s Claude Code Following Controversy
BEIJING, July 3 (Reuters) – In a significant development, Chinese technology behemoth Alibaba has instituted a prohibition against employees utilizing Anthropic’s Claude Code within the workplace.
This directive follows growing concerns regarding the tool’s features, which can potentially identify users linked to China, as disclosed by an anonymous source familiar with the situation.
This ban marks the intensification of tensions between the two entities. Anthropic has accused Alibaba of engaging in unauthorized extraction of its Claude AI model’s capabilities — a dispute that epitomizes the intense competition between the United States and China in the race for artificial intelligence supremacy.
Claude Code, Anthropic’s AI coding aid designed for software developers, has gained traction among programmers in China, notwithstanding the company’s restrictions aimed at curtailing access by Chinese users and entities.
The aforementioned source, who was not authorized to engage with the media and opted for anonymity, indicated that Alibaba employees have been instructed to utilize the company’s proprietary coding platform, Qoder, instead.
Neither Alibaba nor Anthropic has promptly addressed inquiries from Reuters for comment. Notably, Alibaba has refrained from publicly responding to Anthropic’s allegations thus far.
Anthropic Charges Alibaba with Distillation Efforts
In the preceding month, Anthropic claimed that it experienced a “distillation” strike orchestrated by Alibaba, a process involving the training of a less sophisticated model based on the outputs of a more robust one.
This distillation process, Anthropic asserts, accelerates China’s capability to reach its advanced Mythos Preview functionalities, as articulated in correspondence reviewed by Reuters and sent to two U.S. senators.
Alibaba’s prohibition coincides with revelations from developers that Claude Code incorporates mechanisms capable of scrutinizing user environments, including timezone data and proxy-related details, while embedding subtle markers into the prompts dispatched to Anthropic’s servers.
An employee from Anthropic elucidated on X that this feature was a trial initiative launched in March, aimed at safeguarding against account misuse by unauthorized resellers and providing protection against model distillation.
The source that informed Reuters regarding Alibaba’s ban noted that Anthropic’s restrictions targeting Chinese users pose challenges in enforcement.
Individual users, capable of deploying servers in the United States, can obscure their true locations, making it appear as if their traffic originates from within the U.S.
Nonetheless, companies are increasingly cognizant of the legal and compliance risks, the source elaborated.
As developers of AI models in the U.S. strive to thwart unauthorized access, resale, and distillation of their technologies, Chinese cloud and AI enterprises are pivoting towards domestic and open-source alternatives, such as DeepSeek, Alibaba’s Qwen, Moonshot, and Zhipu.
Concurrently, Chinese AI models are beginning to penetrate the U.S. market — a trend that has provoked concerns among certain industry experts within the United States.
Details regarding Alibaba’s ban were initially brought to light by Chinese media outlets.
Source link: Aol.com.
