They claim machine learning, but I'd guess their workhorse is a "grep -f badwords tweets.json | makepdf | send boss@employer" kind of thingy.
E.G the "BIG DICK ENERGY" post being flagged as bigotry and sexism, no doubt they trained on limited data of hand curated "questionable posts" and I wouldn't be surprised if they used a source like 4chan and just automatically assigned negative labels to the vast majority of posts.