Today, Alibaba announced internallyReverse disabling Claude. All Alibaba employees are required to uninstall Anthropic-related products, including Sonnet, Opus, Fable and other series of models, as well asClaude CodeAgent products included. Banned fromJuly 10Officially effective.
According to sources, since the beginning of this year, in order to encourage employees to adopt AI technology, Alibaba has not only introduced free quotas for internal models, but also implemented a large reimbursement policy for the use of external models. Employees canFreely choose Claude, GPT, Gemini and other external models, many programmers consume as much asnumberhundred dollars, and willClaude Code, OpenAI Codex and Alibaba’s Qoder are all frequently used Agent tools.This "reverse disabling" cuts off Claude's channel.
According to the news, the direct trigger of the ban can be traced back to June 24.then,AnthropicIt was revealed that he submitted a letter to the U.S. Senate Banking Committee on June 10,accuse AliBetween April 22 and June 5, approximately 25,000 fake accounts were used to conduct more than 28 million conversation interactions with Claude. This behavior was unilaterally characterized as an "industrial-level model distillation attack" and raised the matter to the level of national security.
In response, Anthropic significantly tightened its risk control strategy. It is worth mentioning that Anthropic has previously used almost the same rhetoric to accuseDeepSeek, Dark Side of the Moon and MiniMax. In February this year, Anthropic publicly issued a document accusing these three Chinese AI laboratories of conducting large-scale distillation attacks on Claude. ("Anthropic accused a Chinese model of "fraud", Musk fiercely attacked him, and overseas netizens criticized him")
The accusation came to light just as Ali sued the U.S. government.On June 24, Alibaba submitted a complaint to the federal court in San Jose, California, suing the U.S. Department of Defense and requesting that it be removed from the "Chinese Military Enterprise List" (1260H List).
The list was released by the U.S. Department of Defense on June 8 and includes Alibaba, Baidu, BYD and many other companies. Although the list is not directly equivalent to formal sanctions, it still raises market concerns about Alibaba's ability to use cutting-edge U.S. technology. Ali later submitted detailed evidence to prove his innocence, but did not receive any reply from the US Department of Defense.
From the end of June to the beginning of July, Anthropic implemented a new round of treatment for Claude.Large-scale account ban, a large number of Chinese users were banned without warning, and both personal subscriptions and team accounts were affected. Among them, accounts that were paid directly through Anthropic's official website and are found to be in violation will not be refunded, and the success rate of appeals is extremely low.
Later, a developer's reverse analysis found that Claude Code had a built-in "hidden Trojan" system starting from version 2.1.91 released in April 2026. The system reads the local time zone, checks whether the proxy or custom API address contains keywords such as Alibaba, Byte and other Chinese cloud vendors, AI companies, API agency service providers, etc., and uses steganography to replace punctuation marks in the system prompt words to mark Chinese users. Claude Code team members later publicly acknowledged this "experimental" measure.