Privacy by default for clinical context.

Do not enter names, medical record numbers, contact details, or other direct identifiers into the workspace.

What you provide

Search questions and optional patient context are sent to this application to retrieve and contextualize evidence for your session. Patient context is used for local applicability and safety checks, and is not included in external language-model requests.

Optional profile memory

Account profile memory is off by default and requires an explicit opt-in under the workspace settings. If enabled, de-identified patient context and recent searches are stored in encrypted, account-scoped application storage for up to 30 days. Evidence results are cached separately for 15 minutes and do not include patient context; the current context is applied locally when results are displayed. You can turn memory off or delete saved profile data from the workspace settings.

What may be sent to the language model

For synthesis, the application sends the search question and selected source passages, abstracts, full text available through PMC, trial summaries, and source metadata to the language-model provider configured for this deployment. The provider and model are deployment configuration, not a fixed or undisclosed vendor. Patient context fields are withheld.

Retention and access

When queued synthesis is enabled, an optional normalized patient context may be stored with the encrypted job for up to one hour. Application-level expiry is enforced on reads; physical removal by the database's expiry process may occur later. Operational infrastructure may retain limited technical logs. Administrators must configure log access, retention, provider retention, and deletion controls appropriate to their environment before production clinical use.

Important limits

This is a privacy design statement, not a guarantee of regulatory compliance. Do not submit identifiable health information. Review the provider agreement, applicable source licenses, and organizational policy before using clinical data.