Click here to visit Original posting
Adding unrelated words, spaces, or 'leetspeak' manages to trick Google's Perspective AI into thinking that abusive messages are in fact harmless, according to research from the Aalto University and the University of Padua.