Shiberu Browser Extension - Privacy Policy
Last updated: 2026-07-05
This policy is written in English; translations are provided for convenience. The English version is authoritative.
This policy explains what the Shiberu browser extension collects, why, where it goes, and what control you have over it. It is written to be read by a real person, but it is complete enough to satisfy Chrome Web Store requirements and the basics of the GDPR.
The extension works alongside your Shiberu account (the app at https://shiberu.app). Data you create in the app is covered by Shiberu's main privacy policy; this document focuses on what the browser extension does.
The short version
•
The extension sends the subtitle text of what you're studying to Shiberu's own servers to make each word clickable and explainable, and to generate AI explanations.
•
It saves the flashcards you choose to save - including an optional screenshot and audio clip - to your Shiberu account.
•
AI explanations and text-to-speech run on Shiberu's own self-hosted models. Your text is not sent to any third-party AI provider such as OpenAI.
•
The shared, cross-user cache that makes Shiberu fast contains no account data - it is keyed by the text of a subtitle line alone.
•
Shiberu only reads subtitle text, never the protected video or audio.
•
Shiberu does not sell your data and runs no remote or obfuscated code.
1. Who we are
Shiberu ("we", "us") provides the Shiberu language-learning app and this browser extension. For any privacy question, or to exercise the rights in section 9, contact
[email protected].
2. What the extension collects and transmits
2.1 Account and authentication data
To save cards and sync your progress you sign in with your Shiberu account. The extension stores your login tokens and basic profile (user ID, email, and your native/target languages) in the browser's local extension storage. Login tokens stay inside the extension's background service worker and are sent only to Shiberu's own backend to authenticate your requests. They are never exposed to the web pages you visit.
2.2 Subtitle and selected text (the core of what the extension does)
When you study a video or read a page with Shiberu, the extension sends the text of the subtitle lines (or the web text you selected) to Shiberu's backend for two things:
•
Analysis / tokenization - splitting the line into words and adding readings, part-of-speech, and dictionary information, so each word is clickable.
•
AI in-context explanation - an on-demand, short explanation of what a word means in that specific sentence, generated when you click a word.
We send the subtitle text itself. We do not send, record, download, or transmit the video or audio stream. On streaming sites we read the caption text the site's own player already loads; we never extract anything from the DRM-protected media (see section 6).
2.3 Word-knowledge progress
The extension keeps track of which words you know, are learning, or want to ignore, so it can color-code them. These per-word statuses (a word and its status, e.g. "learning") are synced to your account so the same progress shows up across your devices and in the app.
2.4 Flashcards you save ("mining")
When you choose to save a subtitle line as a flashcard, the extension sends to your account:
•
the sentence, and optionally its translation, reading, and your notes;
•
context used to label the card: the platform name (e.g. "YouTube"), the show/video title, episode, timestamp, and the page URL;
•
an optional screenshot of the current video frame;
•
an optional short audio clip of the line (a few seconds, capped at 10s), or synthesized speech.
You control this - a card is only sent when you save it. Nothing is captured in the background.
2.5 Preferences and diagnostics
Your settings (display modes, hotkeys, chosen deck) are stored locally and, for some, synced through the browser. The extension may send anonymous adapter-health diagnostics (e.g. "subtitle acquisition on this site fell back to limited mode") so we can detect when a site changes and breaks the extension. These events are about the extension's technical behavior, not about you, and carry no subtitle content.
2.6 What we do NOT collect
•
We do not track your general browsing history.
•
We do not read pages you have not chosen to study or read with Shiberu.
•
We do not access other tabs; the reader and universal player run only on the tab where you invoke Shiberu.
•
We do not capture the protected video or audio itself.
2.7 Diagnostic & telemetry data
To keep the extension working as streaming sites change, it may send anonymized diagnostic events about the health of subtitle acquisition. Each event is anonymous, contains no content, and contains no personal information. An event records only:
•
which streaming platform the extension was running on (e.g. "YouTube");
•
whether it loaded full subtitles or fell back to the limited mode;
•
an error-reason code when acquisition failed; and
These events explicitly contain no subtitle text, no page URLs, and no account identifier - they describe the extension's technical behavior, not you or what you were watching. We use them only to detect when a site breaks the extension and to improve reliability.
3. Where your data goes
All extension network traffic goes to Shiberu's own backend at `https://shiberu.app` over HTTPS. There is exactly one server destination.
AI explanations and text-to-speech run on Shiberu's own self-hosted models, on our own infrastructure. Your subtitle text and words are not sent to OpenAI, Google, Anthropic, or any other third-party AI or translation provider. (Our servers use a standard OpenAI-compatible interface internally to talk to *our own* models - the data does not leave Shiberu's infrastructure.)
We do not use third-party advertising or analytics SDKs in the extension.
4. The shared cache - and why it protects your privacy
To make Shiberu fast and keep AI features free, the results of analyzing a subtitle line and explaining a word are cached and shared across all users. The important part:
These shared caches contain no account data. A cached line-analysis, line translation, AI explanation, or synthesized-audio clip is keyed only by a hash of the text of the line (plus the language). There is no user ID, no account, and no "who watched this" attached to a shared row. If you and a stranger both watch the same episode, you both benefit from the same anonymous cached analysis - but neither of you can be identified from it, and it cannot be traced back to a person.
Your personal data - your word-knowledge statuses and your saved cards - is stored separately, tied to your account, and is never mixed into the shared cache.
5. Data retention
•
Login tokens and preferences: kept in your browser until you log out or remove the extension. Logging out invalidates the session token on the server.
•
Your word statuses and saved cards: kept in your Shiberu account until you delete them or close your account.
•
Shared, anonymous cache entries (line analysis, explanations, translations, audio): retained to serve all users, and periodically pruned. Because they contain no personal data, they are not "your" data in a way that identifies you.
•
Diagnostic events: retained in aggregate to monitor extension health.
Deleting your Shiberu account removes your account-linked data (profile, word statuses, saved cards). Anonymous shared-cache entries, which contain no personal data, are unaffected.
6. Streaming services and DRM
Shiberu is a study tool, not a downloader. On video sites it reads only the subtitle/caption text in order to make words clickable and to let you save a line as a flashcard. It does not decrypt, record, download, or redistribute the protected video or audio. The optional short audio clip saved with a card is captured only on sites where the audio is not DRM-protected; on protected services (e.g. Netflix) no audio is captured and synthesized speech is used instead. Your use of each streaming service remains subject to that service's own terms.
7. Security
•
Login tokens are held only in the extension's background service worker and are never exposed to web pages.
•
All communication with our backend is over HTTPS.
•
The extension contains no remote or obfuscated code; everything it runs ships in the reviewed package, and our backend returns only data (never code).
8. Children
Shiberu is not directed at children under 13 (or the minimum age in your country). We do not knowingly collect data from children.
9. Your rights
Depending on where you live (including under the GDPR in the EU/EEA and the UK), you have the right to access, correct, export, or delete your personal data, and to withdraw consent. You can:
•
edit or delete your saved cards and word progress from within the Shiberu app;
•
log out to stop the extension from sending data;
•
remove the extension to stop all extension processing;
•
email
[email protected] to request access, export, or deletion of your account data, or to ask any privacy question.
Our legal basis for processing is the performance of the service you asked for (making subtitles interactive and saving your cards) and your consent, which you can withdraw at any time by logging out or uninstalling.
10. Data selling and sharing
We do not sell your personal data, and we do not share it with third parties for their own purposes. We do not send your study text to any third-party AI provider. Data is processed only to provide the Shiberu service.
11. Changes to this policy
If we materially change what the extension collects or how it's used, we'll update this page and its "Last updated" date, and where appropriate notify you in the app or extension.