There is a reluctance to discuss at a weight level - this graphs out refusals for criticism of different countries for different models:

https://x.com/xlr8harder/status/1884705342614835573

But the OP’s refusal is occurring at a provider level and is the kind that would intercept even when the model relaxes in longer contexts (which happens for nearly every model).

At a weight level, nearly all alignment lasts only a few pages of context.

But intercepted refusals occur across the context window.

This seems like it may be at the provider level and not at the actual open weights level: https://x.com/xlr8harder/status/1883429991477915803

So a “this Chinese company hosting a model in China is complying with Chinese censorship” and not “this language model is inherently complying with Chinese censorship.”