What happens to what you say

FLOWX turns what you say into clean, ready-to-paste text. This policy describes exactly what that involves: what leaves your phone, who processes it, and how long it is kept. It describes the app as it is actually built, not as we would like it to sound.

What we collect

Your account. An email address and a password. The password is stored only as a bcrypt hash — we cannot read it, and neither can anyone who obtains the database.

Your speech, while it is being transcribed. When you dictate, the recording is sent to our server, transcribed, cleaned up and returned to your keyboard. See “How long audio is kept” below.

What you set up. Dictionary corrections, snippets, your chosen writing style and keyboard layouts. These are the settings that make dictation come out the way you want, so they are stored against your account and apply on every device you sign in on.

Usage totals. For each dictation we record the number of words produced, the number of tokens the AI used, the estimated cost and the time. This is what your weekly word allowance is counted against and what the Stats screen shows.

Diagnostics. The app keeps a log of what it did and when — for example “the microphone was tapped”, “the connection to the server failed, error code -1009”, “the app crashed” — so that we can find and fix problems. The same log is sent to our server automatically every few minutes, together with the app version, the device model (for example “iPhone17,1”), the system version and a random identifier of the installation; when you are signed in it is linked to your account. The log never contains what you say or type: no dictated text, no transcripts, no passwords. We keep it for 30 days, then delete it; it is also deleted with your account (lines sent before you signed in carry only the installation identifier and are deleted after the 30 days). You can turn automatic sending off in Settings → “Send diagnostics automatically”, and you can still send the log by hand with “Send diagnostics log”.

What we do not collect

• We do not take screenshots of your screen, and we do not read the contents of the apps you use FLOWX in.
• We do not record what you type on the FLOWX keyboard. Typed characters go straight into the app you are typing in and never reach our servers.
• We do not sell your data, and we do not use it for advertising.
• We do not use your dictations to train any model, ours or anyone else’s.
• Diagnostics are used only to find and fix problems in FLOWX — never for advertising or profiling, and never shared with anyone.

How long audio is kept

Audio is never written to a database or to disk on our servers. It passes through an in-memory processing queue and is dropped from that queue automatically, and in any case is discarded once the queue reaches its size limit — a matter of minutes under normal use. There is no archive of your recordings and no way for us to play back what you said.

The text produced from a dictation is kept on your device, in your history, until you delete it. Clearing the history in Settings removes it. The text is not stored on our servers.

Your permission comes first

Before your first dictation FLOWX asks, in the app, whether your speech may be sent for transcription, and names the providers below. Until you agree, nothing is recorded and nothing is sent.

Who else processes your speech

To transcribe and clean up your dictation we use these providers:

• Groq — speech-to-text (Whisper) and cleaning up the transcript.
• OpenAI — backup speech-to-text, used only when Groq is unavailable.
• Anthropic — backup clean-up when Groq is unavailable, and edit commands on selected text (the Claude API).

Each receives the audio or transcript needed to do that work and is bound by its own terms as a data processor. At the time of writing, none of these providers trains models on data submitted through their APIs.

Payments

Subscriptions are handled by Apple. We never see your card details. Apple tells our server that a subscription is active; that is all we store about it.

Where data is held

Our servers are in Europe. Our AI providers process data in the United States.

Your choices

• Delete your dictation history — Settings → Clear history, on your device.
• Stop sending diagnostics — Settings → turn off “Send diagnostics automatically”. Nothing more is sent; what was already sent is deleted after 30 days, or at once with your account.
• Delete your account and everything attached to it — Settings → Delete account, in the app. Your account, dictionary, snippets, usage records and the diagnostics linked to your account are erased from our servers immediately, and the history on the device is cleared. It cannot be undone. A Pro subscription is billed by Apple and has to be cancelled separately, in Settings → Apple Account → Subscriptions.
• Stop dictating — nothing is sent anywhere unless you press the microphone. While a Flow session is open the app holds the microphone so the keyboard can reach it instantly; that audio is discarded on the device and never leaves it. Ending the session releases the microphone.

Children

FLOWX is not directed at children under 13 and we do not knowingly collect their data.

Changes

If this policy changes in a way that affects what we do with your data, we will say so in the app before the change takes effect.

Contact

[email protected]