Karchu
Back to documentation

Privacy and data retention

What Karchu stores when you upload a financial document, how long each kind of data is kept, and how retention is configured.

Last updated 2026-09-11

What is stored, separately

  • The original uploaded document
  • The processed representation used for OCR, where one was produced
  • Raw extraction and OCR output, kept so a figure can be traced back to its source
  • Normalised transactions and their provenance
  • Data you create: rules, categories, corrections, comments and reports

Retention

  • The default is to keep records, not to delete them: a processing job finishing is not a reason to discard a financial record
  • Each workspace can set its own retention policy per data kind
  • A purge is previewed before it runs, so you can see exactly what would be removed
  • Deleting a document removes its rows and its stored files

Consent and tracking

  • Analytics and advertising cookies are only set after consent, with GDPR and CCPA choices available
  • Consent choices can be changed at any time from the privacy controls

The legal document is the privacy policy. This page describes how the product behaves.

Questions

Is my statement deleted after processing?
Not automatically. The original is kept so results can be re-checked and reprocessed, and so a reviewer can see where a figure came from. Retention is configurable per workspace, and you can delete a document and its stored files at any time.
Is my financial data used to train a model?
No. Categorization learns through visible, deterministic rules that you approve. There is no automatic model training on your transactions.
Does the assistant send my whole statement to a model?
No. Figures are computed from your own rows in the application. The model is used only to interpret the question and phrase the answer around those computed facts.