Where your text goes

You dictate, the application reads aloud, and an AI rewrites your text. Each of these actions sends text somewhere. Here is where it goes and under which contract, including when no such contract exists.

Updated: August 26, 2026

This English version is a translation provided for convenience. The French version is the binding one. Read the binding French version.

🎤
Your voice becomes text

Your dictated audio is transcribed by a specialized service provider, covered by a signed data processing agreement dated June 24, 2026. The recording is deleted as soon as the transcription is returned; an automated recovery pass follows.

● Signed agreement
🧠
AI processes your text

SpeakApp's AI text processing, including the voice assistant, Reading reformulation, translation, and Dictation transformations, is handled exclusively by Mistral AI, a French company. Requests are sent to its European infrastructure hosted in the European Economic Area.

We have chosen one engine. There is no silent fallback to another vendor: if the service is unavailable, the feature stops and informs you. We have disabled the use of our exchanges to train that provider's models.

● Signed agreement
🔊
Text is read aloud: standard voice

The standard voice is produced by a Microsoft text-to-speech service. This is the one link in the chain not covered by a data processing agreement: the library we use is publicly available and comes with no agreement we can sign. We prefer to disclose this rather than have you discover it later.

In practice, the text SpeakApp reads to you is sent to that service at playback time. We do not retain it.

▲ Disclosed without an agreement
🏢
Text is read aloud: contract-backed voice

For organizations unable to accept the limitation above, a tailored offer is in preparation. It substitutes a synthetic voice supplied by a provider under contract for the standard voice, keeping the text in a fully contractual processing chain. Contact us if you need this arrangement.

● Signed agreement
  • 🔒Your saved snippets, dictionary, and shortcuts stay on your computer and are encrypted there by Windows.
  • 🔒Voice commands are recognized on your computer, without an internet connection.
  • 🔒No AI provider keys to configure: you do not need a third-party account; everything is managed on the server side.
  • 🗄️We do not store the content of your conversations on our servers. Our database tables contain lengths and counters, never conversation text.

These phrases are widespread. We do not use them because we cannot stand behind them.

“100% European”

The standard voice uses Microsoft. Our AI provider has also informed us in writing that some of its own sub-processors may temporarily process data outside the European Economic Area. Those transfers are governed by the European Commission's Standard Contractual Clauses, but they exist.

“No data ever leaves Europe”

For the same reason, and this is not entirely under our control. Even resolving the Microsoft arrangement would not make this statement accurate.

“GDPR compliant” without qualification

Compliance must be audited, not merely declared. We can say that we process data in accordance with the GDPR, with a privacy policy, record of processing, public sub-processor list, and rights that can be exercised at any time.

Why spell out weaknesses? Our sub-processor list is public and a careful reader could find these issues anyway. Disclosing them takes one line; hiding them would cost your trust.

Access, rectification, erasure, portability, and objection rights can be exercised at any time. We handle these requests ourselves.

Write to privacy@speakapp.work. You may also complain to the French data protection authority (CNIL) at any time.