Anti-Scam Protection

Anti-Scam runs continuously in the background — there is no command to trigger it. Every server gets message, URL, and behaviour-based scam detection with automatic deletion and staff alerts. Premium adds AI image scam detection and opt-in cross-server compromised-account protection, sharing intel the moment a hacked account starts spreading scams anywhere in the network.

• Choose a private alert channel and the staff roles allowed to see alerts.
• Decide whether auto-delete should be enabled, or whether staff should review flagged messages first.

1. Open Anti-Scam in the dashboard.
2. Enable message, URL, and behaviour detection, then set the block-confidence threshold and behaviour observation window.
3. Choose the alert channel and staff roles, save, and monitor the first alerts.

• Message, URL, and behaviour-based detection are each independently toggleable and enabled by default.
• A blocked match can auto-delete the message and notify staff in a configured alert channel.
• The block-confidence threshold (default 70) controls how certain a match must be before it is actioned.
• Behaviour detection watches a configurable observation window (default 60 seconds) for scam-like patterns of activity, not just a single message.

Premium adds AI image scam detection, on by default, which catches scam screenshots that text-based filters miss.

• Compromised-account flagging lets this server's detections mark an account as compromised for the whole network.
• Cross-server protection lets this server act on flags other servers raised, with a configurable minimum confidence and a choice of response, from logging only up to a stronger automatic action.
• Cross-server triggers can fire on a flagged message, a flagged account joining this server, or both, and are off by default even on Premium, so each server opts in deliberately.

Anti-Scam is a dedicated, always-on pipeline built specifically for scam messages, links, behaviour, and images. Airwavy AI Moderation is a separate, general-purpose safety classifier covering 14 broader categories, one of which is scams — the two can run together without conflicting.

• A message matching a scam pattern is deleted (if auto-delete is enabled) or flagged, with a staff alert posted where configured.
• Normal messages, links, and activity are unaffected.