We're introducing XAM Guard, a new experimental Discord moderation bot built to help servers respond to image spam attacks faster.
Recently, a lot of Discord servers have been dealing with hacked accounts spamming the same MrBeast-style image over and over. Some servers have been protecting themselves by placing a clear warning channel near the top of the server, then automatically timing out, kicking, or banning anyone who posts there.
The idea is smart: compromised accounts often move through channels from the top down, so the warning channel can catch them before they reach the rest of the server. It works, but it still means staff have to reserve a channel for defense and rely on the spam hitting that channel first.
XAM Guard is our answer to that.
What It Does
XAM Guard watches for suspicious image bursts across your Discord server. When it detects a possible spam event, it sends an alert with the details staff need to act quickly.
The alert can include things like:
- The user who triggered the detection
- Whether XAM Guard believes the content is AI-related spam
- The response mode used
- How many images were involved
- Which channels were affected
- How many matching images were deleted
- Whether a timeout was applied
- How confident the detection is
- The attachments that were scanned
From there, staff can take action directly from the alert, including banning, kicking, removing a timeout, marking the event as a false positive, or simply acknowledging it.
Built for the Current Spam Problem
XAM Guard was inspired by that warning-channel approach, but it tries to catch the same pattern directly instead of making the channel do all the work.
Instead, XAM Guard focuses on the behavior that actually matters: accounts posting repeated image spam across channels in a short period of time.
The goal is simple: flag the burst, show staff what happened, and make it easy to respond before the spam spreads further.
Experimental for Now
XAM Guard is still an experiment. We are testing how reliable the AI detection is in real servers, and we do not want to overpromise before we have more data.
That said, the bot uses a strong reasoning model to help classify suspicious image activity, so we are confident enough to begin testing it publicly and improving it from there.
What Comes Next
We will keep tuning XAM Guard as we learn more from real-world use. False positives, missed detections, and staff feedback will all help shape the next version.
If you want to try XAM Guard or follow development, join our Discord server for updates.