Samoan artists push back on AI song covers misusing language

This story from Samoa is a small but telling example of how generative models can clash with cultural context long before we reach AGI. The issue isn’t just copyright; it’s that current music models treat language as phonemes to be approximated rather than as carriers of identity and respect. When those models are then used at scale for entertainment, they can normalize broken pronunciation and subtly devalue minority languages, especially for younger audiences who may not have strong linguistic grounding.

For the AGI conversation, these frictions matter because they highlight that ‘alignment’ is not only about not killing everyone—it’s also about not casually erasing cultural nuance. As models get more capable at mimicry, communities will demand finer‑grained control over how their languages, accents and musical traditions are used, which could translate into consent mechanisms, regional fine‑tuning or outright blocking in some contexts. Developers who ignore these early signals risk a backlash that could slow deployment of more advanced systems, particularly in smaller markets where trust is fragile.