Privacy policy

Information processed

After you open Reader, the extension processes loaded article text, title, URL, selected passages, article images, your questions, notes and AI answers. In an enabled reading window, new pages are read only within existing site permissions; other sites require your action to enable access.

Reader does not actively read passwords, cookies, form inputs or editable areas, collect unrelated browsing history, follow pagination or crawl entire sites. Page content is untrusted reading material and cannot grant permissions or execute tools.

Standalone mode: your selected model service

Translation, summaries, questions and diagram explanations send the required article context, selections, images, questions and recent discussion directly from the extension to your configured service. OpenAI-compatible Chat Completions and Anthropic-compatible Messages are supported. Images may be converted to data URLs or base64; image-origin access is requested when necessary.

Your API key stays in the extension’s local IndexedDB configuration and is used only to authenticate with your selected model service. It is not sent to Deep Research, page scripts or data exports. Browser storage is not an operating-system keychain. Only a loopback Local Proxy may omit a key; other services require credentials.

Model providers apply their own retention, training and privacy policies. Reader does not promise zero retention by third parties. For local or self-hosted models, processing location depends on that service’s configuration.

Optional Deep Research platform mode

Platform mode requires sign-in and authorization. The URL, title, required bounded article text, selections, section, question and discussion context are sent to the platform, whose AI Engine calls its configured model provider. Diagram questions and image translation upload browser-read PNG, JPEG or WebP bytes when requested. The server does not fetch image URLs. Follow-up questions can use the image, article and recent discussion.

Reading sessions may sync URL, title, page version, selection, discussion, answers and scroll position to the platform database. Only excerpts, notes, AI insights, tags and sources you explicitly save enter the research library. Sync does not store full page archives, image bytes, full-page translation caches or your model API key.

The sign-in service processes account information; the extension stores a platform authorization token locally. Disconnecting requests token revocation and removes the local token, but does not automatically delete previously synced records. The platform and its model provider may retain necessary operational, security or billing records; zero retention is not promised.

Optional web search

Web search is off by default. When enabled, Reader uses the current question and article title to decide whether a search is needed. Translation and summaries do not trigger a search. Only generated keywords are sent to the search service, without full article text, chat history, images, API keys or platform tokens. Avoid private information in web-enabled questions: keywords may include terms from your question or article title.

Reader currently uses Brave/Bing search excerpts; clearly unrelated results are excluded. Standalone mode queries from your browser after you grant search-site permission; platform mode queries from the platform server. The search service receives keywords and network information such as the requesting IP, under its own policies. Excerpts are not full-page verification. Search failures fall back to the original page. Source links and bounded excerpts are saved with local chat history and included when platform session sync is enabled.

Local storage and retention

Sessions, scroll positions, discussions, annotations, saved items and settings remain in extension storage until deleted, cleared or uninstalled. The translation cache keeps up to 500 entries and evicts older entries without a fixed expiry promise. Bounded source and image inputs required by tasks are temporary, cleared when tasks finish or are cancelled, and excluded from reading-data exports.

Exports exclude API keys, temporary task inputs and translation caches. Export anything you need before uninstalling. Installations with different extension IDs do not automatically share local data.

Your controls and deletion

You can deny or revoke site permissions, change services, stop answers, pause or cancel translation, disconnect the platform, delete local sessions or saved items, and clear the cache or all local data. Settings include import and export.

Clearing local data does not delete synced platform records or requests retained by model providers. Research-library items can be deleted on the platform. Contact project support about cloud sessions, account data or other personal-data requests; contact third-party providers through their own channels.

Purpose and sharing

Data is used for user-initiated reading, translation, discussion, saving, sync, security and troubleshooting. Reader does not sell user data or use it for advertising, unrelated analytics, credit assessment or lending decisions. The extension has no separate advertising or behavioral-analytics reporting.

Recipients are your selected model service and, when you enable platform mode, the platform and its model provider. Reader complies with the Chrome Web Store User Data Policy, including Limited Use requirements. Third-party services apply their own policies.

Permissions explained

activeTab and scripting inject article extraction, selection and translation display after you enable reading. tabs identifies the current tab URL, title and window, links evidence to the article, and opens authorization or source pages; it does not scan browsing history. sidePanel provides the native Chrome reading sidebar.

contextMenus provides passage actions and bidirectional original/translation switching. storage stores extension state, platform tokens and temporary task state. unlimitedStorage helps preserve local reading data and bounded caches; it does not imply unlimited growth. alarms wakes tasks you started, without proactively crawling pages.

Fixed host access is limited to techradar.top and www.techradar.top, so a user-initiated “Open in Reader” click can open the source and sidebar. The bridge does not read platform login credentials or unrelated pages. Other HTTP(S) access is optional and requested for the specific article, configured model endpoint or image origin you use. You can deny or revoke access in Chrome.

Contact and policy changes

Project maintainers provide support through public GitHub Issues. Include the extension version, steps and a public test URL. Never post API keys, platform tokens, passwords or private article content. For personal-data requests, first describe the type of request and ask for a secure contact channel.

This policy covers Deep Research Reader. Material changes are reflected here and in the bundled policy, with user authorization where required. The last updated date appears above.