Some AI applications appear to be trying to tell us what we want to hear – and even avoid answers that might upset us. Is AI human after all? If so, should the media be wary?