> ## Documentation Index
> Fetch the complete documentation index at: https://docs.vetox.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Moderation and safety

> How Vetox's moderation layers fit together — Auto-Mod, AI Auto-Mod, Anti-Raid, Protection, moderation tools and logging.

Vetox protects a server in layers. Each one covers a different kind of threat, and they work independently — you can run all of them or just the ones you need.

## The layers

<CardGroup cols={2}>
  <Card title="Auto-Mod" icon="filter" href="/en/moderation/auto-mod">
    Rule-based message filtering: bad words, invites, links, caps, spam, mass mentions. Free.
  </Card>

  <Card title="AI Auto-Mod" icon="sparkles" href="/en/moderation/ai-auto-mod">
    Understands meaning rather than matching text. Catches what keyword rules miss. Black tier.
  </Card>

  <Card title="Anti-Raid" icon="users-slash" href="/en/moderation/anti-raid">
    Detects mass-join floods and screens suspicious accounts at the door. Basic tier.
  </Card>

  <Card title="Protection" icon="lock" href="/en/moderation/protection">
    Limits the damage a compromised moderator or rogue bot can do. Free.
  </Card>

  <Card title="Moderation tools" icon="gavel" href="/en/moderation/moderation-tools">
    Review and lift active bans, mutes, timeouts and temporary roles. Free.
  </Card>

  <Card title="Logs" icon="scroll" href="/en/moderation/logs">
    A record of what happened in your server, across 148 trackable events. Free.
  </Card>
</CardGroup>

## Which layer catches what

| Threat                                            | Handled by                                          |
| ------------------------------------------------- | --------------------------------------------------- |
| Slurs, spam, invite links, caps abuse             | [Auto-Mod](/en/moderation/auto-mod)                 |
| Harassment or scams phrased to dodge word filters | [AI Auto-Mod](/en/moderation/ai-auto-mod)           |
| A wave of accounts joining to raid                | [Anti-Raid](/en/moderation/anti-raid)               |
| Throwaway or bot-like accounts joining            | [Anti-Raid join guard](/en/moderation/anti-raid)    |
| A moderator account being compromised             | [Protection](/en/moderation/protection)             |
| A malicious bot being added                       | [Protection](/en/moderation/protection)             |
| Enforcing rules by hand                           | [Commands](/en/commands/moderation)                 |
| Reviewing or undoing punishments                  | [Moderation tools](/en/moderation/moderation-tools) |

<Note>
  Auto-Mod and AI Auto-Mod are complementary, not alternatives. Auto-Mod is instant and free and catches the obvious cases cheaply. AI Auto-Mod reads intent and catches the rest. Servers on Black commonly run both.
</Note>

## A sensible starting point

<Steps>
  <Step title="Turn on the Auto-Mod rules you need">
    Every rule ships **disabled**. Nothing filters messages until you enable it. Start with the obvious ones for your community and add more over time.
  </Step>

  <Step title="Set up logging early">
    Logging costs nothing and answers 'what actually happened' later. Enable it before you need it, not after.
  </Step>

  <Step title="Add Protection limits">
    Protection is free and is the only layer that helps when a trusted account turns against you. Configure it even on a quiet server.
  </Step>

  <Step title="Enable Anti-Raid if you are publicly listed">
    If your invite is posted anywhere public, you will eventually be raided. Anti-Raid requires Basic.
  </Step>
</Steps>

## Exemptions

Every layer has a way to exclude people and places from enforcement:

| Layer       | Exemptions available                          |
| ----------- | --------------------------------------------- |
| Auto-Mod    | Per-rule disabled channels and disabled roles |
| AI Auto-Mod | Exempt roles, channels and individual members |
| Protection  | Whitelisted roles and users                   |
| Anti-Raid   | Configured per join-guard condition           |

<Warning>
  Exemptions are the most common cause of "moderation is not working". If a rule seems to be ignoring someone, check whether their role is on an exemption list before assuming the rule is broken.
</Warning>