Security basics

Discord Security for Large Servers: 10,000 Members and Up

Past 10,000 members, a Discord server stops being a group chat and becomes a target. A raid there is worth more, a nuke costs more, and thirty staff are thirty phishing targets. The settings match a small server's, but the values change and a few new rules appear. Here are the decisions that matter at scale, with Security Bot settings.

By Sina, Founder, Security BotUpdated (published )9 min read

Staff tiers instead of one moderator role

A single Moderator role with Ban Members on thirty people is a common large-server mistake. Split staff into tiers and give each tier the least it needs:

TierNative permissionsBot commandsAnti-nuke whitelist
OwnerAdministratorAllAlways exempt
Admins (2 to 3)Manage Server, Roles, ChannelsAllChannel and role actions only
Senior moderatorsTimeout Members, Manage Messagesban, softban, kick, mute, clear, lockallNone; limits apply
ModeratorsTimeout Membersmute, clear, lockNone
HelpersNoneNoneNone

Moderators moderate through Security Bot with per-user command permissions, so a compromised moderator has the bot's capped commands rather than native power. Turn on the server's 2FA requirement for moderation. With thirty staff, you cannot check every account by hand.

Verification friction at scale

Large public servers get hundreds of joins a day, and every extra step costs some real members. The answer is not less verification but smarter verification:

  • Adaptive Verification on Relaxed or Balanced for the default join path. Established accounts are in at once, and it tightens during join spikes.
  • The web check with VPN and alt blocking for servers that are permanently targeted (crypto, NFT, marketplaces).
  • Anti-raid's account age bypass (Premium) for accounts older than an age you choose, such as a year. Established accounts then never add to the raid score.
  • Verification level Medium rather than High, unless you are under sustained attack. The ten-minute wait at High can cost you newcomers on a big server.

Raid scoring for high join volume

A launch, a stream, or a collab can bring a thousand legitimate joins in an hour. Anti-raid (Premium) has to tell that apart from a raid.

Know how it acts. Each join adds points for the signals you enable:

  • A young account.
  • Joined seconds after someone else.
  • No profile picture.
  • The normal-join score, when nothing else matched.

When the running total reaches your score limit, it applies the punishment to every account that added points during the current scoring window, not only the raiders. It also locks the channels you picked.

So on servers with frequent spikes:

  • Keep the score limit high. The default is 200, and every clean join adds 2 points. A limit of 40 is reached by 20 ordinary joins inside the window, and everyone who joined in it is punished.
  • Turn on the account age bypass.
  • Choose Mute (a timeout) rather than Ban as the punishment, so a real newcomer caught in the window can be released with /unmute.
  • Lock only general chat channels. Leave announcements, support, and staff channels off the list.

If you run another moderation bot that mass-bans, whitelist it for the ban action (Users: Ban) and keep its role below Security Bot's. Otherwise its bulk bans count toward your ban limit and anti-nuke punishes it.

Example: launch day with 3,000 joins

This is an illustration, not a real server. A game studio announces a beta and links the server from a trailer. Three thousand joins arrive in ninety minutes, with a raid group riding the wave. Here is how a setup tuned for spikes behaves, and where it can still catch real members.

  1. Anti-raid has the account age bypass on for accounts older than a year. Those joins are recorded with no points and are never punished.
  2. Younger accounts that join seconds apart add account-age and join-row points. Two hundred accounts created that morning push the total over the limit. Anti-raid times out every account that added points in the current window, which includes some real newcomers. It locks the general channels for the lock duration you set (20 minutes by default). Announcements and support stay open.
  3. One-click verification keeps the real newcomers moving. The verification channel is the only one unverified members can see, so nothing gets posted by anyone who has not clicked.
  4. The security log shows the Anti Raid Lock. The moderation log lists each timeout and channel lock with the reason "Anti Raid", and the owner gets a DM. The channels reopen on their own when the lock duration ends, with their previous Send Messages setting. Staff can also reopen them early with /unlock or /unlockall.

The next morning, review who was timed out and release real members with /unmute. If too many were caught, raise the score limit or lower the minimum account age for the bypass rather than turning scoring off.

Nuke limits for busy mod teams

While Anti Ban is on (the default), moderators who ban through /ban are capped at the ban limit per moderator per 10 minutes. Further commands are refused until the counter resets. A team cleaning up fifty scam accounts a day will hit a ban limit of two mid-cleanup.

Raise the limit to what a single moderator does in ten minutes during a bad hour (the default is 5). Rely on Beast Mode for the server-wide picture. Keep channel and role delete limits low. A staff member rarely needs to delete several channels in ten minutes without telling anyone.

  1. Ban limit 5 to 8, kick limit 5, punishment clear roles.
  2. Channel and role delete limit 2 or 3, punishment ban.
  3. Beast Mode score limit 15 to 20, with action scores such as 4 points for a channel delete and 2 for a ban (every action score starts at 0).
  4. Whitelist the admin role for channel and role actions only.
  5. Recovery turned on (Premium; dashboard switch or /recovery enable), and a restore tested once so staff know how it works.

These are suggested values for a busy team, not defaults.

Whitelist hygiene at scale

The whitelist is where large servers quietly lose protection. Every entry is an exemption, and an All entry exempts that role from every anti-nuke check. With thirty staff and a dozen bots, the list grows until it covers exactly the accounts most likely to be phished.

  • Whitelist by role, per action: the admin role for channel and role actions, never for Ban or Bot-Add.
  • Whitelist utility bots for the one action they perform, for example a ticket bot for channel create.
  • Keep Allow whitelist bypass off in Beast Mode unless every whitelisted account has 2FA and you accept the risk.
  • Review monthly on the Whitelists page and remove staff who left.

There is no limit on the number of entries. The dashboard takes up to 100 users per category, so whitelist roles rather than long lists of people. Full walkthrough: how to whitelist bots and users in Discord.

Extra rules for crypto, NFT, and creator servers

  • Threat Link Guard on, with ban as the follow-up (Premium). Wallet drainers are a constant threat in these communities.
  • A member profile rule for names containing your project name, Support, Admin, Mod, or Team, exempting the real staff role.
  • Pinned scam notice in every public channel: staff never DM first, there is no mint link in DMs, official links live in one channel.
  • Announcement channels read-only with Mention Everyone limited to one role.
  • Manage Webhooks removed from everyone, and Anti Webhook Create on (Premium).

Server networks: one policy, several servers

A network of three or seven servers has the same attack surface multiplied, plus one new risk: a staff account that is trusted everywhere. Keep the policy identical across servers so nobody has to remember which one is the exception. Keep whitelists per server rather than copying an All exemption around.

  • Premium per server. Premium slots come in packs: one server at $4.99, three at $12.99, seven at $24.99 per month, assigned from your account dashboard. Premium runs on a separate bot: after assigning a slot, add the Premium Bot to that server. Settings from the free bot do not carry over.
  • Premium+ for networks that need their own identity. A custom bot with your own name, avatar, and status. Plans run from one bot on one server up to four bots on eight servers, from $14.99 per month.
  • One staff member per time zone per server holds LockAll, and a Recovery restore has been tested once on each server.
  • The same log layout everywhere, so a moderator covering another server knows where to look.

Logging at scale

At 10,000+ members, one log channel for everything becomes noise. Split it:

  • The moderation log for staff actions through the bot (plus anti-raid timeouts and locks and AutoMod punishments).
  • The security log for anti-nuke warnings and punishments and anti-raid locks.
  • On Premium, the AutoMod log for filters.
  • On Premium, an audit log channel with only the event packs you actually review (bans, role changes, channel changes).

Logs are messages in your own channels, so keep them in a private staff category.

Common mistakes on large servers and how to fix them

  • One Moderator role with native Ban Members. Thirty accounts can each wipe the member list. Move bans behind the bot with per-user command permissions on the Moderation page.
  • Bot role below the admin roles. Everyone above Security Bot is exempt from every trigger. Drag the role to the top, and check it again after every role reshuffle.
  • Anti-nuke limits copied from a small server. A ban limit of two refuses moderators' /ban commands mid-cleanup and punishes native bans at the second one. So someone raises it to fifty. Use 5 to 8 with clear roles, and Beast Mode for the server-wide view.
  • Verification level High on a public server. New members wait ten minutes and some leave. Use Medium plus Adaptive Verification, which tightens on its own under attack.
  • One log channel for everything. Nobody reads a channel that scrolls a hundred lines an hour. Split the logs and trim audit packs to what you review.
  • No drill. The first restore happens during the incident. With Recovery on, delete a test channel and restore it from the Recently deleted list once a quarter. Also replay an attack in the Beast Mode simulator.

Incident plan for a large server

  1. Two named staff per time zone hold the LockAll and UnlockAll command permissions.
  2. One pinned staff message: how to lock, how to pause invites, where the log is, how to restore.
  3. A quarterly drill: run the Beast Mode simulator and restore a deleted test channel with Recovery.
  4. After every incident, one paragraph in the staff channel on what changed.

The order of operations for the incident itself is in how to recover a Discord server after a nuke. The full setup order is in the setup guide.

Do this with Security Bot

Discord Anti Nuke Bot: Stop nukes with per-user action limits, automatic punishments, owner alerts, and a security log.

Security Bot is a Discord anti nuke bot that watches bans, kicks, channel and role changes, bot adds, prunes, and dangerous permission edits. When an account hits a limit, it is kicked, banned, or stripped of its roles. The server owner gets a DM.

Frequently asked questions

What security does a large Discord server need?

Staff tiers with least privilege, 2FA required for moderation, verification tuned for low friction, anti-raid scoring tuned for join spikes, anti-nuke limits sized for a busy team, Beast Mode with action scores set, split log channels, and Recovery turned on before an incident.

Should a large server use captcha verification?

Use Adaptive Verification as the default. Established accounts get straight in, new accounts get a check, and it tightens during join spikes. Crypto and NFT servers under constant pressure can use the web check with VPN and alt blocking.

What anti-nuke limits work for a big moderation team?

A common starting point is a ban limit of 5 to 8 with clear roles, and a channel and role delete limit of 2 or 3 with ban. Add a Beast Mode score limit of 15 to 20 with action scores set. Adjust after a week of reading the security log.

How do web3 servers stop wallet drainers?

A maintained scam-domain filter with ban as the follow-up, a profile rule for staff impersonation, pinned scam notices, and official links kept in one channel.

Does Security Bot work on large servers?

Security Bot has been added to more than 1,000,000 Discord servers since 2019. On a large server the settings matter most: size anti-nuke limits for your team, tune anti-raid for join spikes, and split the log channels.

How many whitelist entries does a large server need?

As few as possible. There is no entry limit on any plan, and the dashboard takes up to 100 users per category. Use per-action role entries and one entry per utility bot, and review the list monthly.

Should a server network run one bot or several?

Premium on each server (3 servers $12.99, 7 servers $24.99 per month) keeps one policy. Once a server's slot is assigned, add the Premium Bot there. Premium+ adds custom-named bot instances if the servers need separate identities.

How this guide was made: drafted with AI assistance, then checked against Security Bot's source code and Discord's official help articles. Spotted something out of date? Tell us in the support server.