On our servers
By default the check runs on our servers. No third-party AI for scam images or default AI moderation.
Check on our servers in Europe
AI Moderator
Detect swearing and toxic messages on our servers in Europe, the same check as in Discord. You choose how strict; OpenRouter is optional for the whole platform.
What does it do?
AI Moderator checks messages for toxic tone, swearing, hate and threats. The same check as in Discord, on our servers by default, or optional OpenRouter if you turn that on. You choose how strict and what happens: delete, channel notice, warn, or log only.
By default we do not send your chat to third-party AI. Scam image scans always stay on our servers.
Live demo
Test the AI check
Type a test message, see what the same check as your Discord bot thinks.
Toxicity score:
Scored only, not stored, not shared.
Max 10 tests per minute
Why in SecureKy
NL & EN
Works in NL and EN: tone plus rules for swearing, hate and threats.
Staff decides
Mild / normal / strict, delete, channel notice, warn, timeout and mod-log.
Same as Discord
This demo uses the same check as in Discord.
OCR
Optional toxicity in screenshots via shared OCR with anti-scam.
OpenRouter optional
Optional for the whole platform: local or cloud AI. The dashboard shows which is active.
Under the hood
- Local tone check (or optional OpenRouter) for how toxic a message sounds.
- Rule signals: swearing, hate, threats, sexual spam.
- Score from 0 to 1; action only if the check triggers and the score is high enough for your strictness.
- Images: one OCR pass shared with anti-scam when both modules scan.
Settings
| On / off | Module Active in the dashboard. |
|---|---|
| Mild | Threshold 0.85, fewer false positives. |
| Normal | Threshold 0.70, recommended. |
| Strict | Threshold 0.55, catches more edge cases. |
| Delete | Remove the message on violation. |
| Channel notice | Short notice in the same channel. |
| Warn | Via warnings module (optional). |
| Log channel | Embed with score for mods. |
FAQ
Do messages go to OpenAI or Google?
Not for the default check on our servers, and not for scam image scans. OpenRouter is an explicit choice; then only the text for the check goes to that AI. The demo follows the same choice.
Is the site demo the same as Discord?
Yes. The same check and strictness as in Discord. Pick Mild/Normal/Strict to see when the bot would act.
Does it work in Dutch?
Yes, NL and EN. English works too.
Many false positives?
Start on Normal. Too strict? Switch to Mild. Staff/admins are skipped by default.
Does AI Moderator cost extra?
No, part of the free core. The default check runs on our servers; OpenRouter uses separate models if you enable it.