Back to features
AI moderation
Keyword filters match strings; people break rules with meaning. AI moderation reads each message and scores it against thirteen categories, then fires your rule through the same pipeline as any other automod rule.
What you get
- Thirteen categories, each with its own sensitivity
- Per-category actions, from an alert to a ban
- Short messages, commands and raid spam never leave your server
- Fails open: an outage never blocks your chat
How it works
Pick how sure the model must be before it acts, per category: Very Low only catches near-certain violations, Medium is the default most communities want, and Custom sets an exact threshold. Every preset is free on every plan.
Scope it like any automod rule: only public channels, everywhere except a vent channel, or with certain roles exempt. Start with alert-only and read the log before you let it act on its own.
Nighthawk only keeps a per-server count of checks, never the content or who wrote it, and message text is not used to train models. The free allowance covers most communities; DeepIntel raises it for one server and adds image scanning.