Pseudonymization and AI

How your data is masked before sending, and how to verify everything.

How pseudonymization works

Three local detection layers, stable tokens, and a mapping table that never leaves your machine.

Review detections before sending

Dismiss a detection, catch a missed term, and guarantee that the approved preview is what leaves.

Choosing an AI model

Managed models with your account, your own API keys, or a fully local model with Ollama.

Verify what goes over the network

Send badge, exact payload, audit log and token probe: trust is meant to be verified.

The local reading sheet

A structured analysis of your documents, produced by a model running on your machine, no network involved.

The six confidentiality levels

What you declare before sending, what each notch means, and the two that change how the app behaves.

Can’t find the answer?

Write to us: we reply quickly, in French or in English.

Contact us