SlopSieve — Privacy Policy

Effective date: 2026-04-19
Contact: thomas.friedel@gmail.com

Summary

SlopSieve runs its main scoring pipeline entirely on your device. No
post content, profile data, or browsing activity is transmitted as
part of ordinary scoring.

Two categories of features do send data to servers operated by the
SlopSieve author — but only when you explicitly invoke them via a
right-click menu or by pressing a flag/label button. Those features
and what they send are listed below.

What the extension does

SlopSieve is a browser extension that flags likely bot, AI-generated,
or low-quality posts on X (twitter.com / x.com), Reddit (reddit.com,
including old.reddit.com), and Hacker News (news.ycombinator.com). It
reads publicly-visible post and author data that your browser is
already loading, runs a local machine-learning model entirely on your
device, and displays a verdict badge next to posts.

Data read from pages you visit

To function, the extension reads the following kinds of data from
pages you visit on supported sites:

• Public post content (text, language, engagement counts — tweets on
  X, submissions and comments on Reddit and Hacker News)
• Public profile metadata of the author (display name, handle,
  follower/following counts on X, account age / karma where available,
  verified status, country and language settings, default-profile
  flag)

This data is used for local scoring. It is not transmitted off
your device as part of the automatic scoring flow.

Data stored on your device

The extension stores the following locally in your browser, using the
storage API:

• Your settings (filter mode, enabled/disabled state, theme)
• A short-lived in-memory cache of recent scoring results, to avoid
  re-scoring the same account repeatedly. This cache is cleared when
  the browser or service worker restarts.

We do not use cookies. We do not use analytics. We do not fingerprint
your device.

Data transmitted off your device

The extension does not make any remote network calls as part of its
automatic scoring flow. It does, however, include three opt-in
features that transmit data only when you explicitly invoke them:

1. "Score selection with SlopSieve" (right-click menu). When you
   highlight text on any webpage and choose this menu item, SlopSieve
   opens a new tab at https://slopsieve.com/score?text=... with
   your selected text (up to 1500 characters) as a URL parameter, so
   the slopsieve.com web app can run a prediction on it. Nothing is
   sent unless you click this menu item.
2. "Inspect URL with SlopSieve" (right-click menu). When you
   choose this menu item, SlopSieve opens a new tab at
   https://slopsieve.com/inspect?url=...&title=... with the URL and
   page title of the current tab. Nothing is sent unless you click
   this menu item.
3. Account flagging / labeling. If you press a flag or label
   button in the extension's UI (e.g., "Flag this account for
   review"), SlopSieve sends the account identifier, your chosen
   label, and any optional reason text you typed to
   https://api.slopsieve.com. This is used to improve the model
   over time. Nothing is sent unless you press a flag/label button.

Aside from these three explicit user-initiated flows, the extension
makes no remote network calls of its own. It does not contact any
third party. It does not send analytics, telemetry, or heartbeats.

The extension does observe X's own GraphQL traffic on pages you are
already viewing (via interceptor.js) in order to obtain richer tweet
data than the DOM exposes. This observation is read-only and never
forwarded anywhere.

Permissions and why they are needed

• storage — save your settings and a short-lived result cache.
• alarms — periodic housekeeping of the local cache.
• host permissions (*://x.com/*, *://twitter.com/*,
  *://*.reddit.com/*, *://news.ycombinator.com/*) — required to
  read public post content on pages you visit, so the on-device model
  can score them, and to render verdict badges in-page.

See the store listing for a detailed per-permission justification.

Children

The extension is not directed at children under 13 and does not
knowingly process data from them.

Changes to this policy

If this policy changes, the new version will be published at the same
URL and the "Effective date" above will be updated.

Contact

Questions or concerns: thomas.friedel@gmail.com