A back-end server system for step-up authentication for conversational interfaces using spoken passphrases receive a conversational request for an action which requires step-up authorization; respond to the user through conversational artificial intelligence (AI) warning the user that an additional step is required; transmit, over a dedicated communication channel to the user and different than the conversational interface, a secret passphrase generated using a spoken passphrase dictionary; receive, by the conversational interface, a spoken submission from the user in response to the warning; confirm validity of the received spoken submission by comparing the spoken submission to the passphrase; and in response to confirmation of the spoken submission, authorize the requested action.