This category looks for hate speech and discriminatory slurs in context — not only dictionary slurs. It works alongside the instant slur path.
What it catches
- Slurs and dehumanizing language toward protected groups.
- Clear expressions of hatred or contempt tied to identity.
- Some coded or indirect phrasing when the model read is confident.
Intent softening
When the message looks like an obvious joke, meme, or quote of someone else's words, a harsh read can be softened one step (block becomes review, review becomes pass) for this category and a few similar ones. Hate that targets people seriously is not excused by a casual lol.