Alibaba Bans Claude Code as Its Feud With Anthropic Goes Public

Alibaba has told its staff to stop using Claude Code, Anthropic's popular AI programming tool, escalating a feud between the two companies that has spilled from a quiet technical dispute into an open corporate rift.

The Chinese technology giant has classified Claude Code as high-risk software and set a ban that takes effect on July 10, according to reporting from TechCrunch and CNBC. Employees are being pointed instead toward Qoder, Alibaba's own coding assistant.

A hidden line of code

The immediate trigger, security researchers say, was the discovery of code inside the tool that appeared to check whether a user was based in China or linked to a Chinese AI lab. To an Alibaba engineer, that reads like a tripwire. To Anthropic, it was housekeeping gone stale.

Thariq Shihipar, who works on the issue at Anthropic, addressed it directly on X. The check, he wrote, was "an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation." He added that the team "landed stronger mitigations since then and we've actually been meaning to take this down for a while", an admission that the offending code was both old and, by Anthropic's own account, overdue for removal.

What the fight is really about

Behind the ban sits a heavier accusation. Anthropic has alleged that Alibaba and its Qwen AI lab carried out what it called the largest known "distillation attack" against it, generating roughly 28.8 million exchanges with Claude through about 25,000 fraudulent accounts between April 22 and June 5.

Distillation is a well-worn technique: you train a new model on the outputs of a stronger one, so the smaller "student" learns from a "teacher" without needing the teacher's original training data. It is legitimate when a company distills its own models. It becomes contentious when the teacher belongs to a rival and the terms of service say no.

Alibaba has not offered a public, on-the-record explanation for the ban, and its engineers' security concerns and Anthropic's abuse concerns are, in a sense, two sides of the same coin. Each company suspects the other of overreaching, and each has a plausible story.

The bigger picture

The episode is a snapshot of a widening split. American labs are increasingly wary of Chinese firms training on their outputs; Chinese firms are increasingly wary of American software running on their machines. Coding assistants sit right in the crossfire, because they see everything a developer types.

For now the practical fallout is contained to Alibaba's own workforce and its push toward home-grown tools. But the trust question it raises is not going away. When the software writing your code can quietly ask where you are sitting, the line between anti-abuse measure and surveillance becomes a matter of who is telling the story.