【Node: US-China AI Moves Toward Separation】

Alibaba today asked all employees to uninstall Claude, effective July 10.
Add to that Anthropic’s recent complaint to the Senate accusing Alibaba of industrial distillation, of blocking Chinese users’ accounts, and that it was reportedly uncovered for secretly tagging Chinese domain names—
and most people treat it as gossip about two companies ripping their relationship apart.
In fact, this is also a landmark moment showing US-China AI has fully shifted from “competing on the same field” to “separate arenas with isolation.”
Two years ago, the US-China AI circles were still reading each other’s papers and using each other’s open-source models. They argued, but at least they were still in the same technical arena.
Today, one side takes a business issue to Congress, elevates it to a matter of national security, embeds code in its own tools specifically to identify Chinese users, and imposes large-scale account bans. On the other side, it directly uproots the other side’s product from the entire R&D chain.
From the accusations against DeepSeek in February to today’s Alibaba’s reverse crackdown and takedown—within just a few months—the AI world has been split into two halves.
On one side are OpenAI, Anthropic, and Google; on the other side are Qwen, DeepSeek, and ByteDance.
Two sets of models, two sets of standards—no one dares to rely on the other anymore.