A while ago I tested questions about Tienanmen Square and Tank Man with various open Chinese LLMs and one or two closed ones, I think GLM 5.2. None of them refused to answer or blocked the keywords. They gave responses politically aligned with China. Abliterated models gave proper responses.