One way to bypass the censor is to ask it to return the response by using numbers for alphabets where it can. e.g. 4 for A, 3 for e etc.

Somebody in reddit discovered this technique. https://www.reddit.com/r/OpenAI/comments/1ibtgc5/someone_tri...