devplacepy/devplacepy/templates/docs/content-moderation.html

<div class="docs-content" data-render>
# How moderation works

This page describes, without omission, the safety controls DevPlace operates: what is filtered, how
you report, what happens next, and how long it takes.

## The commitment

**Every report is reviewed within {{ moderation_sla_hours() }} hours.** That is not a slogan: the
moderation queue carries a badge showing the age of the oldest unresolved report against that window,
so a breach is visible to every administrator the moment it happens.

## The three mechanisms

### 1. Automated filtering, at the moment content is created

Every piece of user-authored text passes through a classifier before it is published, at the five
places where content is created or changed: content creation, content editing, comment creation and
editing, direct messages, and profile and signup fields. Five choke points cover every surface,
because creation on DevPlace already funnels through them.

The classifier returns one of four verdicts:

| Verdict | Effect |
|---|---|
| `allow` | Nothing happens. |
| `label` | The item is published and carries an age label. |
| `review` | The item is published **and** a report is raised for a moderator. |
| `block` | The item is refused with an explanation. |

The default mode is `review`, not `block`. This is deliberate. DevPlace is a developer platform: its
members discuss exploits, cryptography, malware analysis and violent subject matter as their work, so
a machine that silently suppresses them would destroy the product. A flagged item is therefore
published and a human looks at it. Only the sexual and exploitative categories block outright,
because there is no legitimate developer case for them and the harm of publishing is immediate.

**The filter fails towards review, never towards publication.** If classification errors, the content
is published and a `critical` report is raised naming the failure. A safety control that fails
silently is worse than none.

### 2. Reporting, on every surface

Every surface that can carry user content carries a **Report** control: posts, comments, gists,
projects, project files, news, uploaded media, direct messages, quizzes, polls, awards, profiles,
issue reports, workspaces and assistant output. Select a reason from the
[Community Guidelines](/docs/community-guidelines.html) list, add detail, submit.

Reports are private. The reported person is never told who reported them. You are acknowledged
immediately and can follow the outcome at `/reports/mine`.

Filing a second report on the same item updates your existing one rather than creating a duplicate. A
different person reporting the same item creates a separate report, and the queue shows the count, so
a widely-reported item rises to the top.

### 3. The queue and the decisions

Administrators work one queue, sorted oldest-unresolved-first. Each report shows the reported item in
place, who authored it, how many people reported it, and every previous decision against that author.

A decision is one of:

| Decision | What it does |
|---|---|
| Remove content | Removes the item from every surface |
| Restore content | Reverses a removal |
| Warn | Records a warning against the account and tells the user |
| Suspend | Blocks creating content for a stated period, with a stated reason |
| Ban | Closes the account permanently and revokes every credential |
| Lift | Ends a suspension or ban |
| Dismiss | The report was not a violation |
| Escalate | Raises severity and keeps the report open for a second opinion |

Removal is not available for targets that have no removal path - direct messages, accounts,
workspaces, polls and assistant output. For those, the remedy is the account-level action.

Every decision writes a permanent record: what was decided, by whom, why, and against which report.
Every decision is also written to the append-only audit log. **The user is always told what was
decided and why**, through an in-app notification, and can reply to dispute it.

Administrators cannot act against an administrator who registered before them.

## What a suspension actually does

A suspended account can still sign in, still read everything, still see the reason for the
suspension, and still delete itself. It cannot create or change content. A user is never trapped.

## Blocking, which is yours alone

Independently of reporting, you can **block** any account. Their content disappears from your view
everywhere except their own profile, they can no longer message you, and they can no longer notify
you. It is one-directional, reversible, and never disclosed to them. **Mute** is the lighter version:
it stops their notifications while leaving their content visible. See
[Block and mute](/docs/block-and-mute.html).

## Age labelling

Content classified as mature is hidden behind an interstitial by default. Revealing it is an explicit
choice, recorded on your account, and it is not offered at all to accounts in a minor age band.

## If you disagree with a decision

Reply to the notification, or write to the address on the [Contact](/docs/contact.html) page. State
the report and what you believe was wrong. A different administrator reviews it.
</div>