Quick story: we turned on AI moderation last Tuesday for our developer community (~1,200 members). Set it to flag-only mode, not auto-hide, because we wanted to see how it performed before trusting it.
Within 48 hours it flagged a post that looked totally normal on the surface — a "helpful" code snippet in a reply that actually contained an obfuscated XSS payload. Our mod team had approved it manually because the surrounding text was legit.
The AI flagged it under the Security Risk category with a note about the encoded script tag. Impressive.
Our settings
| Setting | Value |
|---|---|
| Categories | All 6 enabled |
| Threshold | Medium |
| Action | Flag for review (not auto-hide) |
| Review cadence | Daily |
Anyone else have good AI moderation stories?