What happens to a file you upload.

1Where files are processed

Application servers

Railway, region EU West (Netherlands). The files you upload are stored and edited here.

Web and API gateway

Vercel, region Frankfurt (Germany). Uploads pass through on their way to the application servers.

Account database

Supabase, Frankfurt (Germany). Accounts, plan and usage counts — no document content.

Error reports

Sentry, EU data region. Technical error reports only — no document content.

Payments

Stripe. Billing details only — never documents.

2How long files are kept

Uploaded files and results

Deleted automatically 2 hours after your last activity, and 24 hours after upload at the most. Translations are kept with the batch and deleted with it.

Originals

Never altered. Every change is written to a new copy.

Trial counting

A fingerprint (hash) of each file is kept to count distinct documents — not the file or its content.

3Where AI is — and isn’t — used

Editing

Find & replace, masking, formatting, pages, stamps and metadata are done by deterministic code. No AI model sees your documents.

The assistant

Reads only what you type — your instruction — never the documents. It runs on Anthropic’s Claude API, so your instruction text is sent there.

Translate (Uses AI)

Only when you run it, the text of the selected documents is sent to Anthropic’s Claude API to be translated.

Anthropic’s handling

Per Anthropic’s published policy: API data is not used to train models by default; inputs and outputs are deleted within 30 days (kept up to 2 years only if flagged under its usage policy); data is stored in the US and may be processed in the US, Europe, Asia or Australia.

Anthropic Privacy Center

4Access and transport

In transit

HTTPS only, with HSTS.

Sign-in

Google sign-in. A batch is tied to the account that created it; other accounts can’t reach it.

Training

Acropea never uses your documents to train anything.