Trust at VoiceIt

The short version

• Your voice never leaves your Mac. Speech-to-text runs on-device, every time, no exceptions.
• AI is opt-in. Plain dictation never calls any AI service. Only Prompt Mode and Agent Mode do, and only after you press the hotkey.
• AI processing stays in Europe. When you do use AI, your transcript is processed by Google Vertex AI in Frankfurt, Germany — never the US.
• No training on your data. Your transcripts are not used to train any AI model.
• No tracking, no ads, no analytics. We have no idea what you dictate. We don't want to.

What stays on your device

Everything below is processed entirely on your Mac. We never see it. No server, no log, no backup.

• The audio captured from your microphone while you hold a hotkey
• The transcribed text produced by the on-device speech recognition model
• Your transcription history (kept for up to 30 days, then auto-deleted)
• Your custom dictionary, voice fillers, and personal prompt templates
• Your app settings (hotkeys, language, theme, microphone selection)
• The 30-day audio archive (auto-purged on a rolling basis)

If you uninstall VoiceIt and remove ~/Library/Application Support/com.newways.voiceit/, all of it is gone.

What we transmit, and only when you opt in

Prompt Mode and Agent Mode (AI text processing)

When you trigger Prompt Mode or Agent Mode, the already-transcribed text (and in Agent Mode, the text you had selected in another app) is sent to a large language model for restructuring or transformation.

• Audio is never sent. Only the transcribed text.
• Routing: request goes to our Supabase Edge Function (voiceit-proxy-vertex), which calls Google Vertex AI in europe-west3 (Frankfurt). If Frankfurt is unavailable, traffic fails over to europe-west4 (Amsterdam). Both are EU regions.
• Model: Google Gemini 2.5 Flash on Vertex AI.
• Data sent: the transcript, your selected text (Agent Mode only), the system prompt, your anonymous device ID for rate-limiting, your subscription tier.
• Data retained: none on our side. Google retains a short-lived abuse cache (~24 hours) as standard practice for Vertex AI.
• Training: Vertex AI does not train on customer data. This is the default under the Google Cloud Data Processing Addendum.
• Opt-out: simply don't use Prompt Mode or Agent Mode. Plain Dictation never calls any AI service.

Update checks

VoiceIt fetches latest.json from github.com/newwaysai/voiceit-releases at startup and periodically. Standard HTTPS request, includes only what GitHub logs by default (IP, User-Agent).

License verification

If you have a paid subscription, the app contacts Supabase at startup and every 30 minutes to confirm your subscription is active. Data sent: anonymous device ID.

Crash and error reports

Errors are logged only to your local machine (~/Library/Logs/com.newways.voiceit/). No automatic crash reports are sent anywhere.

In-app feedback

If you choose to send feedback via the in-app form, that text plus your email (if you provide one), device ID, tier, app version, and OS family go to our Supabase project (EU, Dublin/Ireland). Nothing else.

Where AI processes happen

All AI processing happens in the EU. The only US sub-processors involved in VoiceIt are Stripe (payments) and GitHub (static file hosting for app updates) — neither of them ever sees your dictation content.

What we never do

• ❌ We never send your audio recordings anywhere
• ❌ We never train any AI model on your data
• ❌ We never run analytics or telemetry inside the app
• ❌ We never sell or share your data with advertisers
• ❌ We never embed third-party tracking scripts in the app
• ❌ We never read your transcription history — it stays on your Mac
• ❌ We never route AI traffic through the US

Sub-processors at a glance

Full Data Processing Agreements with all sub-processors that handle personal data. Listed individually in our Privacy Policy.

Your rights under GDPR

You have the right to:

• Access the personal data we hold about you (Art. 15)
• Correct inaccurate data (Art. 16)
• Delete your data — right to be forgotten (Art. 17)
• Restrict processing (Art. 18)
• Port your data to another service (Art. 20)
• Object to processing based on legitimate interests (Art. 21)
• Withdraw consent at any time (Art. 7(3))
• Complain to the Berliner Beauftragte für Datenschutz und Informationsfreiheit (datenschutz-berlin.de)

To exercise any of these rights, email marcel@newways.ai. We respond within 30 days.

How VoiceIt compares

Last verified: 2026-05-22. We re-check competitor claims quarterly.

Read more

• Privacy Policy (English) — full GDPR documentation
• Datenschutzerklärung (Deutsch) — Original German version, legally binding
• Terms of Service
• AGB
• Imprint / Impressum

For deeper questions about how we handle data, write to marcel@newways.ai.