Perhaps not, but who knows what kind of spaghetti code cascading effect purposely limiting and censoring massive amounts of sensitive topics could have upon other seemingly completely un-related topics such as math.

For example, what if it's trained to recognize someone slipping "N" as a dog whistle for the Horrific and Forbidden N-word, and the letter N is used as a variable in some math equation?

I'm not an expert in the field and only have rudimentary programming knowledge and maybe a few hours worth of research into the topic of ai in general but I definitely think its a possibility.