That sounds exactly like what is going on. Are there specific ways of prompting it that can let you manipulate that moderation classifier?
(I almost think half of the point of opening up CharGPT like this is to flex the fact that they can prevent it from being Milkshake Ducked by the usual level of cherrypicking and malevolent use.)
It is worth noting that one of OpenAI’s public product is a moderation classifier running its own (continuously updated) model and providing scores in various “objectionable content” categories; it’s possible they are using something like a more advanced version of this to determine whether, and what kind, of “I won’t respond because…” answer to use, rather than something relying only on manual identification of particularly narrow content.