My interpretation is it quickly generates answers to keep it conversational but another process parses those messages for "prohibited" terms. Whether that secon

My interpretation is it quickly generates answers to keep it conversational but another process parses those messages for "prohibited" terms. Whether that second process is automated or human-powered is TBD