ShipLedger AI is intended for trade and logistics documents that often contain confidential commercial information, including invoices, packing lists, air waybills, bills of lading, customer identities, supplier identities, pricing, and in some cases banking details.
Based on the current codebase, uploaded extraction documents are accepted into server memory and passed through the extraction pipeline without being written by the extraction route to the application database or permanent file storage. The browser-side extraction result is then kept only in the active session for review.
Registered account details such as business email, account metadata, selected plan interest, and workspace setup information may be retained as part of account creation and customer onboarding. That category is separate from the temporary handling of one-time extraction files.
Separate from one-time extraction uploads, the app includes features for saved templates and saved company profiles. Those items are intentionally retained when a user chooses to save them. They are not part of the “non-persistent extraction” promise and should be disclosed separately.
The extraction service in this codebase sends submitted document content to Anthropic Claude for analysis. That means users are not only sending files to the ShipLedger service, but also to an external AI processor as part of the extraction workflow. Any production disclosure should make that explicit.
If a user uploads a confidential invoice, packing list, AWB, or BL, the current app appears designed not to retain that extraction document in its own permanent storage during the extraction step. However, the document is still transmitted for processing, and saved templates or saved profile data are a different category of information that the app may retain when the user chooses to store them.