If it can be man made, it can be man broken as the old saying goes. So no censor can be absolute - just look at some of the flagging problems we have here on Steemit.
But this Summer a new coded lexicon has appeared on Twitter from these alt-wing groups (ultra right wing groups). They are famous for using the triple parentheses earlier in the Summer to denote someone of Jewish origin e.g. [[[ ]]] - Yes, I do indeed have some Jewish roots (by blood) myself - my grandmother was Jewish. But, because the Twitter algorithm views these parentheses as harmless, the alt-wing groups can communicate their hatred online freely.
Another systematic abuse is the use of everyday words: googles, yahoos and skypes all mean something derogatory. I recently read one that said "Gas the googles, chain the yahoos." - in this instance googles meant the n-word and yahoos meant spics. Skittles is also a new code-word for Muslims since the Donald Trump refugee incident.
Trolls are becoming even better at defeating the Google algorithms that search for hate, and perhaps even Steemit has its own version too - though I am unsure of this??
Only machine learning from Google can help eliminate this use of this wording - by knowing which context the word is being used in. Was the word I used in the last sentence a google or Google??
Until then, this new lexicon will change and mutate, so beware, there be trolls out there!