Alibaba Bans Employees from Using Anthropic's Claude Code Amid AI Dispute

Alibaba has instructed its employees to cease using Anthropic's Claude Code for work, signaling an escalation in the ongoing dispute between the two artificial intelligence firms.

The directive, first reported by Chinese media and confirmed by sources familiar with the matter, stems from Anthropic's accusation that Alibaba has been illicitly extracting the capabilities of its Claude models. Anthropic alleges that Alibaba employed a technique called "distillation" to train a smaller system on the outputs of its more powerful models, aiming to rapidly advance its own AI development.

Anthropic detailed its concerns in a letter to two U.S. senators, stating that Alibaba's actions were intended to accelerate its progress toward matching Anthropic's advanced Mythos Preview capabilities. The dispute intensified recently when developers discovered that Claude Code inspects user environments, including timezone and proxy settings, and embeds subtle markers into prompts sent to Anthropic's servers. An Anthropic employee confirmed this functionality was an "experiment" launched in March to prevent account abuse and protect against model distillation.

In response to the alleged misuse, Alibaba employees are now being directed to use the company's proprietary coding platform, Qoder, instead. While enforcing restrictions on individual users routing traffic through U.S. servers can be challenging, the source noted that enterprises face greater legal and compliance risks. Neither Alibaba nor Anthropic has publicly commented on the allegations. This conflict unfolds against a backdrop of increasing competition and tension between the U.S. and China in the global AI landscape, with Chinese companies like Alibaba increasingly turning to domestic and open-source AI models.

“

an experiment we launched in March