The new prompt detects “potentially harmful or offensive” replies to a tweet before a person hits send, according to a Twitter statement Wednesday, the same day it rolled out the new feature across iOS and Android accounts, starting with accounts that have enabled English-language settings.
