Even if you use it on Chinese servers, the data they collect is likely no more or less than what any other AI company does.

So the questions is about the threat model - how likely or possible it is to use that data against you?

I think it's much better than what was available before OpenAI/Google. I can run it locally and the data are not shared with any Chinese or USA entity. Not even hitting the internet cables.

(1) If you run the model on your server there is no problem

(2) They have a privacy policy, so it boils down to (a) do you like it? and (b) do you trust them to follow it?

They have to follow Chinese law.