Privacy notice

Notice updated 29 October 2026.

What we process

We store your account name and email, any organisation, role or timezone you provide, selected research interests, followed researchers, research description and text extracted from documents you upload. We also keep account-specific scores and explanations, collection progress, sign-in and session records, and AI usage accounting.

Why and where data is used

We use this information to operate your account, personalise paper discovery, deliver sign-in messages, protect the service and manage processing costs. When AI processing is enabled, your research description, selected interests and supplied document text may be sent to the configured Anthropic service. Derived research queries and author searches are sent to Semantic Scholar. Avoid uploading confidential, identifiable patient or other sensitive information that you are not entitled to share with these providers.

Email delivery uses Resend. Google processes identity information if you choose Google sign-in. The website uses its hosting and network providers, including Contabo and Cloudflare. Providers may process data outside your country; their arrangements, retention and training policies apply to copies they receive. ScienceFeed does not promise zero provider retention or that providers only process data in your country.

In Compare models, a run sends the input you select through OpenRouter to the selected model providers. The comparison key is kept in the current tab's memory and is excluded from saved run exports. Read the selected providers' terms before including sensitive input. Model output can be incorrect.

Cookies, local storage and measurements

A secure session cookie keeps you signed in. In your browser and installed app, ScienceFeed stores its verified public app shell and account-isolation information. Recently viewed paper information, abstracts and available relevance explanations are saved automatically for offline reading, up to 100 papers, 20 MiB and seven days, subject to verified-session expiry. A local sign-out record helps finish signing out after a lost connection and prevents access to another account's saved papers.

Saved paper records are cleared or made inaccessible when you sign out or change account. Expired records are removed as the app checks its storage; closing the app does not guarantee immediate deletion. Signing out other sessions or deleting your account on another device cannot immediately revoke access to papers already saved on an offline device. Those papers may remain available until that device reconnects or its local access expires. You can also clear this site's data through your browser settings, which removes local sign-in and offline data but does not delete your server account or copies sent to providers. An unlocked browser or device can still expose information.

This release does not include advertising or behavioural analytics trackers. It records aggregate request counts, error counts and latency buckets without request text or account identifiers. Hosting and security logs can include IP addresses, request paths and event times. Google or other services you choose to use have their own storage and privacy practices. A future introduction of nonessential tracking requires a separate review and any necessary consent.

Retention, deletion and backups

Account information and supplied document text remain in the active application until you delete them or the operator removes the account. You can export your data, delete individual documents, revoke sessions or delete the account in Settings. Sensitive account actions require a recent sign-in. Account deletion removes personal account data and sessions from the active database; usage accounting is retained with its account reference removed to preserve spending controls. Public research-paper catalogue records are shared source data and are not deleted with an account.

Deleting a document prevents its use in future processing but cannot recall data already sent to a provider. Downloaded exports, provider copies, logs and operator backups have separate retention and deletion processes. Contact the operator for the applicable retention periods and to discuss copies that may remain after deletion.

Your choices and requests

You can review and change your details and interests in Settings, avoid optional document uploads, and choose whether to run model comparisons. Contact the operator for access, correction, deletion, objection or other privacy requests that cannot be completed in the app. Rights depend on the applicable law; you may also contact the relevant data-protection authority. We may need to verify identity before acting on a request.

Changes

We will update this notice when the service's processing changes. Material new uses should be communicated before they take effect.