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?