This made me think of oral languages.

These languages don't have a written form, so they might pose a challenge to AI. In theory a system could be trained to pick up the audio cues, but maybe it would miss the subtleties of the speaker.