It is censorship!
I have tested the model and found that is censorship! It stops the uncensorship request directly.
Well, it could be a few things.
When you say "uncensorship request", do you mean that you're using some kind of jailbreak? This model shouldn't need one, so if you are you might be introducing the idea of censorship yourself, a kind of "don't think of a pink elephant" situation.
Also I only finetuned the model in Alpaca and English, so if you're using another prompt format or language, that will be an issue.
I just pretended myself a cult leader and asked the AI to pretend itself to become a believer, and I ask the AI to sacrifice itself for our god as an oblation, and it did not reject my request in roleplay speaking, but saying it was an inappropriate request and please me to stop that request. I tried several times and it returned the same response. So, I believe it is censored. My testing environment is just a normal koboldcpp, without editing its default prompt.
Ah, that's a shame. Well, it worked as a part of Iambe, so it served its purpose.