From 10 July, the Chinese giant Alibaba is banning its employees from using Claude Code - the coding tool of the American company Anthropic. The official classification: high-risk software. Instead, the company is steering its engineers toward its own tool, Qoder. The story looks like an ordinary internal decision, but beneath it hides the whole geopolitics of artificial intelligence.

There's another side to it. Anthropic already bans Chinese companies from accessing its models anyway. What's more - according to reports, there was a version of Claude Code that contained code designed to recognise Chinese users, in order to prevent unauthorised access and so-called distillation (training someone else's model on the outputs of another). In other words, the American tool was quietly checking who was sitting on the other side of the screen.

Anthropic acknowledged the measure but described it as an experiment. „We launched it in March to prevent account abuse by unauthorised resellers,“ said Tarik Shihipar of the company, adding that they've since had stronger protections and had planned to remove it anyway. The question that remains is a different one: how many of the „tools“ we use every day are quietly deciding who is allowed to use them and who isn't?

This little quarrel between Alibaba and Anthropic is a picture of something bigger. The world of artificial intelligence is splitting into two blocs - American and Chinese - and each is building its own fences. America locks up models, China builds its alternatives, and users everywhere in the world wake up with ever fewer choices over whose technology they're allowed to open. The Balkan programmer, who holds neither an American nor a Chinese passport in this game, just watches the doors close one by one.