RefractrRefractr logo

Security and data handling

What happens to the documents you send Refractr, in plain terms. Written for the IT and security people who have to approve it. Last updated 8 November 2026.

In short

Where your data goes

Your system → HTTPS → Cloudflare (network edge) → Refractr, the Netherlands → response

The extraction model is Refractr's own and runs on our own machines in the Netherlands. There is no third-party AI or cloud processing of document content.

Requests pass through Cloudflare's network for encryption in transit and protection against abuse. Cloudflare does not store document content.

Retention

DataKept for
Document, schema and extraction resultNot stored. Processed to answer your request. An asynchronous result is held for a limited time so you can retrieve it.
Same, with Help improve Refractr switched onStored and used for training. Personal data is removed or pseudonymized from training examples.
Usage metadata (timestamps, job IDs, sizes, status, IP addresses for rate limiting)As long as needed for billing, support and abuse prevention
Account and billing recordsAs long as the account is active, plus legal and accounting retention
Database backupsEncrypted, on our own hardware in the Netherlands

Training is off by default for new accounts. Once an example has been used to train a model, it cannot be removed from that model.

Training

Access and accounts

Subprocessors

SubprocessorPurposeSees document content?
CloudflareNetwork edge, encryption in transit, abuse protectionIn transit only, not stored
StripePayments for credit purchasesNo
AhrefsCookie-free statistics on public web pagesNo
Zoho (EU)Email for account messages and supportOnly what you email us yourself

Reviewing Refractr for your company? Email us your questions and we will answer within one working day.

← Back to Refractr  ·  Privacy Policy  ·  Terms of Service