But. That's just me, my pessimism-sci-fi scenario.
But still, the most recent version of american foss model gpt-oss is just so filled with censorship that its just not worth it in the name of "safety", so to me both are doing censorship but I'd much rather use chinese censorship since its only censored on chinese topics and I mean, I personally wouldn't be ever asking chinese models chinese questions but maybe that's just me.
And even if I would, I would probably ask it on some uncensored, in fact I was actually thinking of creating a fine tune like the perplexity, or just this idea to break chinese censorship.
I also think that some better idea needs to come up with multi modal approach so that censorship could be removed by mixing and matching american and chinese models, I do think it is far from reality but I read a recent comment about harmony which gpt-oss uses and it does look promising I am not sure.
And models can be abliterated to remove the censorship. I use llama3 that way.