Exclusive reports indicate that Alibaba (BABA) has internally announced a comprehensive prohibition on the use of Claude. All employees are required to uninstall products related to Anthropic, including the Sonnet, Opus, and Fable series of models, as well as agent products like Claude Code. This ban is set to take full effect on July 10.
According to informed sources, since the beginning of this year, Alibaba had encouraged staff to adopt AI technologies by offering free internal model credits and substantial reimbursement policies for using external models. Employees were free to choose from external models like Claude, GPT, and Gemini, with many programmers reportedly incurring weekly costs of several hundred dollars. They frequently used agent tools such as Claude Code and OpenAI Codex alongside Alibaba's own Qoder. This new "reverse ban" specifically severs the access channel for Claude.
The immediate catalyst for the prohibition appears to be an incident reported on June 24. At that time, it was revealed that Anthropic had submitted a letter to the U.S. Senate Banking Committee on June 10. The letter alleged that Alibaba had used approximately 25,000 fake accounts to engage in over 28 million conversational interactions with Claude between April 22 and June 5. Anthropic unilaterally characterized this activity as an "industrial-scale model distillation attack" and elevated the issue to a national security level.
In response, Anthropic significantly tightened its risk control policies. It is noteworthy that Anthropic had previously used nearly identical language to accuse other Chinese AI labs, including DeepSeek, Moonshot AI, and MiniMax, of conducting similar large-scale distillation attacks on Claude earlier this year.
The timing of these allegations coincided with legal actions taken by Alibaba against the U.S. government. On June 24, Alibaba filed a lawsuit in the U.S. District Court for the Northern District of California, seeking its removal from the "Chinese Military Companies" list (the 1260H list) published by the U.S. Department of Defense on June 8. The list included several companies such as Alibaba, Baidu, and BYD. While not constituting formal sanctions, the listing raised concerns in the market about Alibaba's continued access to advanced U.S. technology. Alibaba submitted detailed evidence to prove its non-military affiliation but reportedly received no response from the U.S. Department of Defense.
From late June to early July, Anthropic implemented a new wave of large-scale account suspensions for Claude, affecting a significant number of Chinese users without prior warning. Both individual subscriptions and team accounts were impacted. Accounts deemed in violation that had paid directly through the Anthropic website were not refunded, and the success rate for appeals was reportedly very low.
Subsequent reverse engineering analysis by developers suggested that Claude Code, starting from version 2.1.91 released in April, contained a concealed system. This system reportedly checked the local time zone and scanned for keywords related to Chinese cloud providers, AI companies, and API proxy services—such as Alibaba and ByteDance—within proxy or custom API addresses. It also allegedly marked Chinese users through steganographic techniques that replaced punctuation in system prompts. Members of the Claude Code team later publicly acknowledged these "experimental" measures.
Comments