STOP SLOP privacy policy

Effective date: 8 September 2026. Applies to extension version 0.8.2.
Contact: STOP SLOP support.

Automatic checks

STOP SLOP checks supported feed posts and opened articles as you browse. It starts in Mark mode with cloud checks enabled for supported feeds. Opened articles receive local checks until you enable cloud for that exact website in Settings. You can pause scanning or turn cloud checks off in Settings. The display slider changes how flagged content appears; Review mode still runs automatic checks.

Local checks run in your browser. For eligible visible feed content and articles on websites you have approved, the extension sends the extracted text, a platform code and a random installation identifier to the STOP SLOP service on Cloudflare. The service normalizes that text and sends it through OpenRouter to the DeepSeek V4 Flash Latest route. Each installation receives 5,000 free fresh checks per UTC calendar month. We keep the service key on the server. After exhaustion, you can add your own OpenRouter key in Settings. The extension then sends automatic checks directly to OpenRouter using the same DeepSeek route, with charges to your OpenRouter account.

Automatic requests contain up to 8,000 characters from the start of a post or article. We do not add the author handle, page URL, DOM, cookies or browsing-history log. Text itself can identify someone or contain private information. The extension excludes recognized messaging and utility pages and requires an article container for general websites. These checks cannot guarantee that all page text is public. Turn cloud checks off before browsing content you do not want processed by the service.

Cloudflare receives your IP address and normal connection information. We use Cloudflare's IP rate limiter to control abuse. OpenRouter and its provider receive submitted text and the service's connection information. We request no provider training and zero data retention through OpenRouter's routing controls. Provider availability and third-party policies still apply. See OpenRouter's privacy policy and Cloudflare's privacy policy.

Caching and retention

We keep text hashes, bounded review labels and reason codes, timestamps in Cloudflare Durable Object storage for 30 days. We also retain an Unknown result for interrupted or ambiguous failed attempts to avoid resending text. If the service rejects a request before admitting a model call, the browser keeps a shorter cooldown and may send it after that pause. A daily cleanup removes expired records, so deletion can take up to one extra day; Cloudflare recovery snapshots may retain metadata under its platform policy. Our application does not persist or log the raw post text. We disable Worker application observability for the scan service.

The extension keeps up to 1,500 hash/result cache entries for 30 days and display settings in local browser storage. We share cached server results for identical normalized text across users and platforms. Hashes do not guarantee anonymity: someone who knows the text can compute a matching hash. The cache contains no separate author identity or page URL. We do not use these requests or cache records to train a model.

The extension keeps a random installation UUID and cached monthly allowance in extension-origin IndexedDB. It sends the UUID to our service to enforce the free allowance. We store a hash of the identifier with the month, usage and reset time; we do not forward it to OpenRouter. We remove expired monthly records through bounded daily cleanup. Older extension versions use a hashed IP bucket for allowance accounting. These identifiers are pseudonymous, not proof of a unique person. Reinstalling or clearing extension storage can create a new identifier. Shared cached checks do not consume a model allowance; fresh reservations can count even when the result is Unknown or a dispatched attempt fails.

Your browser storage and controls

The extension stores display settings, approved article website origins, account exceptions, community lists and subscription URLs, prior list versions and text fingerprints for personal exceptions in your browser. Account exceptions currently support X. They express your preferences and do not verify authorship. You can remove these entries in Settings. Uninstalling removes extension storage from that browser profile; it does not delete the shared server cache.

If you add your own OpenRouter key, the extension keeps it and your personal review allowance in extension session storage. Page scripts cannot access that credential through the content scripts. Closing the browser session or reloading the extension clears the session key. Automatic checks after the free monthly allowance and manual reviews share your adjustable personal session limit. OpenRouter receives your key and normal network information, including your IP address, when you use this direct route. We do not receive your personal key.

Advanced manual text and image reviews

You choose whether to send a manual text or image preview by clicking its review button. These requests go from the extension to OpenRouter and its selected provider using your own key. They share the personal session allowance with automatic checks made using your key and may incur charges on your OpenRouter account. OpenRouter receives your key and normal network information, including your IP address.

Before image conversion, we inspect the original file on your device for embedded C2PA Content Credentials. This check does not validate signatures or invisible watermarks. A profile sample contains eligible loaded post text; the extension does not add the profile biography, handle or URL. Image review prepares a resized JPEG preview and removes the source filename and EXIF metadata. Visible identifying information inside an image remains visible. Selecting a file alone does not submit it. We request no-training and zero-retention endpoints for these reviews too.

Topic catalog, community publishers and research exports

The optional AI-topic filter matches multilingual phrases and model and publisher names on your device. The extension includes a catalog and checks the STOP SLOP service for an update at most once per day. This request sends no post text or browsing history. We retain the last working catalog if an update fails. The bundled community starter needs no network request; its updates arrive with extension releases.

Subscribing to a community list grants access to the HTTPS publisher you select. The extension fetches its list and checks for updates once per day. The publisher receives connection information, but the extension does not send browsing history or post text with list updates.

Community proposals and research contributions export local files for you to inspect and share. We provide no contribution upload endpoint. Exporting does not submit a sample for training. An imported experimental classifier runs on your device in observation mode and keeps session aggregate measurements. It cannot hide posts or skip actual review requests. We do not receive its observations.

Website and support

This website uses no analytics scripts or advertising trackers. Your browser sends connection information to Cloudflare when loading pages or media.

The support form sends your reply email, message, optional name and optional screenshot to the STOP SLOP team through Cloudflare Email. Selecting a screenshot creates a browser preview; submitting sends the file. Remove private content, metadata and credentials before sending. We do not add your browsing history or OpenRouter key to support requests.

Cloudflare Turnstile processes browser and connection information to protect the form. We verify its token on our server and use your IP address for request throttling. Our application does not store support submissions in a database or log their contents. We receive correspondence in our support mailbox and use it to respond. Contact support to request deletion. Email providers may retain records under their policies. We do not use support messages or screenshots for training.

Data use and questions

We use data to provide the filtering, review and support features described here. We do not sell it, use it for advertising, or use it for creditworthiness or lending decisions. We do not collect analytics or clickstream reports. Authorship estimates can be wrong; Unknown does not trigger AI-writing filtering. Your topic and community filters can still match that content.

Contact support for privacy questions or data requests. Your rights may depend on your jurisdiction. Provide enough information to identify your request without sending credentials or unrelated sensitive content. We will update the effective date when these practices change.