Servers on Recommended version 1 keep the old limit of 6 until someone adopts version 2 on the preset update card or in /setup › Settings › Presets.
Новое
The Defendr website is live at defendr.live: sign in with Discord and manage your server's Defendr settings in panels in your browser, with the same rights you have in Discord.
The website is available in 11 languages, and status.defendr.live shows whether Defendr is running.
Исправлено
If Discord's message data leaves out a member's roles, Defendr now looks the member up before it acts, instead of treating an admin like any other member.
If that fails, Defendr deletes only clear threats such as known scam links, scam images or dangerous files and alerts staff; the member is never punished, messaged or named in public.
In Recommended version 2, file spam starts at 10 messages with files within 10 seconds instead of 6; a post counts once, however many files it has.
Monitor-only version 2 has the same value, but the rule stays off there.
Strict stays at 4 messages with files. New servers start on version 2, and servers with "Auto-adopt safe preset updates" switched on move to it by themselves.
The privacy policy, terms, imprint, accessibility statement and security policy (version 2026-10-14) now also cover the website.
If you limited /config, /security, /verify, /updates, /antinuke or /raidmode to certain roles, check /setup and /lockdown in Server Settings › Integrations: those limits do not move with them.
If you limited /case (or the old report menus) to certain roles, set the same limit on /warn, /timeout, /kick, /ban and Report in Server Settings › Integrations: new commands do not inherit it.
Новое
Members now have five commands: /defendr (status, stats, changelog, invite, support, your data, legal texts), /check, /report, /appeal and /help.
Admins manage everything in /setup, the new server control center with Find a setting: protection, raid, anti-nuke, verification, logs and updates, settings.
Moderators get /case with a user card and its actions, the shortcuts /warn, /timeout, /kick and /ban, and /lockdown as one emergency panel that also runs raid mode.
The message menus Report to Defendr and Report to moderators are now one menu, Report. Check user shows the user card with its actions.
Defendr staff use one command: /ops opens the ops center by task (queues, look up, intel and feeds, system, team and server).
An old command answers with where it moved, and /help › What moved in 2.1? lists every one. These aids stay until release 2.2; the new panels show a New in 2.1 line for 30 days.
Исправлено
/lockdown now shows in the command list for members with Timeout Members; an admin with only Manage Server no longer sees it there. Locking and raid mode stay admin-only.
The data export (/defendr › My data › Export) still allows one copy per day, counted only after the whole copy arrived.
Anti-nuke now counts bans and kicks a moderator has Defendr carry out (/case, /ban, /kick, the join card, warn escalation) like the moderator's own: same limits, same trusted lists.
Ban these accounts on the raid card and Ban executor on the anti-nuke card now need your own Ban Members permission.
Nothing is lost: settings, cases, appeals, reports and data requests carry over unchanged.
The groundwork for the Defendr website is included and stays switched off until the website is announced.
The privacy policy and the terms (version 2026-10-13) name the new places: /defendr › My data, /defendr › Support and /defendr › Changelog.
Moderators can act through Defendr only on members below their own highest role, as in Discord: /case and the Ban and Kick buttons on join cards refuse anyone else, administrators included.
Revert, False positive and overturning an appeal lift a member's timeout only when your highest role is above theirs, as in Discord. Lifting a ban works as before.
Nobody but the owner acts on the server owner through Defendr. If Defendr cannot check the role order at that moment, it does nothing and asks you to try again.
Name patterns in the join gate are checked in a way that always finishes at once. Certain patterns could make Defendr pause for seconds on every join, on every server.
The privacy policy in Defendr now matches the published version of 8 October 2026 (the defendr.live notice page).
Defendr no longer stalls for seconds to minutes once an hour: the hourly upkeep of the threat lists now gives way to the bot between short steps and leaves what remains for its next run.
While a threat feed is held back for an unusual jump, Defendr no longer processes the same list again in full every 15 minutes.
The nightly clean-ups give way to the bot between their steps as well.
A network listing the owner accepted alone now becomes active by itself after 72 hours if its evidence is still there and no appeal or restriction stands. It is a hint while listings are hint-only.
The few manual entries carried over from Defendr 1.1 that were still waiting for a review become active as the owner's decision. They stay hints, and each listed person is told and can appeal.
The owner can switch this automation off at any time; until then every automatic step is reported to the Defendr team.
Исправлено
The daily network reminder no longer repeats listings that Defendr confirms by itself; listings that fail a check are still named every day.
The join gate works again in servers with * or ? in a name pattern. Until now it did nothing there (age, avatar, names, action), and every join ended in an error.
ScammerAlert entries become active network hints at once, marked as a third-party list Defendr did not check. They stay a warning for moderators and never act by themselves.
When a listed member joins, moderators get a card in the joins log; ScammerAlert's stated reason shows only in a staff-only channel. Ban or Kick on a known-scam card is their choice.
Исправлено
An entry ScammerAlert strikes through or deletes is withdrawn automatically; a corrected entry is taken over again.
Moving between the two lists updates the listing, and only a struck account ID counts as a withdrawal.
After a restart or an outage Defendr starts checking the whole list again within minutes (over several runs, at most once an hour), and every processed message in #scam-api gets a receipt.
A new ScammerAlert entry for a person whose earlier listing the Defendr team closed waits for a person at Defendr instead of becoming active.
"Show ScammerAlert hints" (the former "Show unreviewed hints") now switches off every ScammerAlert hint of a server, active ones included.
A listed person can also appeal a ScammerAlert listing to the Defendr team, not only at ScammerAlert.
A failing lease renewal ends a job run after a few attempts, and the threat-feed stream backs off longer when a connection keeps dropping.
Hijack detection now needs evidence: a post repeated across channels is at most deleted, unless it carries a known scam, spreads a risky link or invite at script speed or shows a clear attack pattern.
Server owners, administrators and members with a staff role are never timed out, messaged or named in public because of a multi-channel post; staff is alerted instead.
Исправлено
One gallery post or a list of allowed links no longer counts as attachment or link spam.
@everyone inside code or a quote and the ping of a reply no longer count against the author.
Invites posted by apps and webhooks are logged instead of deleted, and bots are no longer treated as hijacked accounts unless they post a known scam.
An established member's first single spam hit only deletes the message; a repeat within 10 minutes keeps the configured timeout.
Repeats of one incident are folded into its open case instead of opening new ones, and the member gets at most one recovery message.
The compromised mark follows only a timeout that applied and ends when a moderator lifts that timeout.
Defendr no longer restarts when a network connection or the database is briefly unavailable.
A nuke tells the whole story, and the owner always hears about it
Нужны действия
Set a log channel with /setup and confirm your trusted list with /antinuke trust, so Defendr can act against an attack on its own instead of only alerting.
Новое
An anti-nuke alert now lists everything the attacker did (channels deleted, members kicked, apps added and more), not only what crossed a limit.
If Defendr cannot show an anti-nuke alert in your server, the server owner gets one direct message with the full record and a Restore button.
Only the current server owner can use that button.
Исправлено
The privacy policy (version 2026-09-26) describes this message to server owners.
Run /setup once to choose a preset and a logging channel.
Launch week: Defendr removes scams but does not time out, kick or ban yet. Sanctions switch on for all servers once it looks clean. To only watch, pick Monitor-only in /setup.
Set an alert role under /config logging if cases that need a human should ping someone.
Set an updates channel with /updates channel to receive release notes here.
Новое
Link and media protection now uses reviewed threat intelligence with per-server actions you control.
Text inside images is checked like any other message, so a scam posted as a screenshot no longer slips through.
Hijacked-account protection stops members whose account posts the same scam in several channels.
Raid detection opens an incident, offers a one-click lockdown and restores every channel afterwards.
Every enforcement gets a numbered case with its steps, a Why? explanation and a false-positive button.
Members can appeal any sanction with /appeal; staff decide it in a private queue, and a wrong case is reverted.
Presets and overlays replace hand-tuning: pick Monitor-only, Recommended or Strict and adjust single settings.
All protection is free; there is no paywalled safety feature.
Исправлено
Content a member did not write - bot replies, webhooks, link previews, forwarded messages - can never get them sanctioned.
Repeated spam from the same member folds into the open case instead of opening a new one for every message.
A case that needs a human is marked on its log card and mentions your alert role once, instead of going unnoticed.
New indicators from threat feeds only watch for 24 hours before they are allowed to act.
A logging channel can no longer point at a channel that belongs to another server.
Update posts are per server and localized instead of one global broadcast.
Report and appeal buttons are authorized on every click, not only when the card is built.