Privacy Policy
Last updated: 13 August 2026
1. Who is responsible
The data controller for Kaho (the desktop app) and wovero.com (the website) is the legal entity named in the Imprint. Contact: privacy@wovero.com.
2. What we collect
When you use Kaho, we process:
- Audio recordings — captured from your microphone while you hold the dictation hotkey. Sent over TLS to our API for transcription. Not retained after the transcript is returned.
- Screen context — the name and title of the app you're dictating into (e.g. "Slack — #general") plus the visible text in the field at the cursor. Used to adapt tone, fix vocabulary, and apply translation. No screenshots, no full-screen capture, no pixels.
- Account data — email address, name, and OAuth identifier provided by Clerk during sign-in. Stored on our backend so you can sign in across devices.
- Usage counters — number of words transcribed per month per user, for free-tier enforcement and cost estimation. No transcript text is retained.
- Glossary & profile — your custom vocabulary, role, writing style, bio, and custom rewrite rules. Stored to personalize transcription quality. Visible to you via the desktop app's Settings.
3. What we don't collect
- We do not store your audio recordings on our servers
- We do not take screenshots or read your screen content beyond the focused text field
- We do not access your files, documents, or browsing history
- We do not sell your data or share it with advertisers
- We do not use your transcripts to train AI models
4. Sub-processors
To deliver Kaho's transcription, cleanup and translation features, we process your data with the processors below. By default, your speech is transcribed on our own GPU hosted in Germany (asr.wovero.com) — US providers are only involved if you explicitly enable cloud fallback in the app.
| Processor | Purpose | Location | Data sent |
|---|---|---|---|
| Wovero GPU (on Hetzner) | Default speech-to-text transcription and LLM cleanup, on our own GPU (asr.wovero.com) |
Germany (EU) | Audio waveform + transcribed text. Discarded after transcription; no retention. |
| Groq, Inc. (opt-in fallback) | Fallback audio-to-text (Whisper-Large-V3 Turbo) and fast LLM cleanup (Llama 3.3 70B) — only when you enable cloud fallback | USA | Audio waveform + transcribed text. No retention. |
| OpenAI, L.L.C. (opt-in fallback) | Reasoning-grade rewrite/reshape and translation (GPT-4o, GPT-4o-mini) — only when you enable cloud fallback | USA | Selected text + your profile + glossary. No retention; not used for training (zero-retention API tier). |
| Clerk Inc. | Authentication and user identity | USA | Email, OAuth identifier, session metadata. |
| Hetzner Online GmbH | Website and API hosting — our own virtual machines run nginx and the API gateway | Germany (EU) | Encrypted in transit; account/glossary metadata stored at rest. |
| Plausible Analytics | Cookieless, privacy-friendly website analytics — no cookies, no cross-site tracking | EU-hosted | Aggregated page views and referrers only. No cookies, no personal identifiers. |
| Resend, Inc. | Transactional email (invites, sign-in links) | USA | Email address, invite content. |
5. Cross-border transfer (Art. 44 GDPR)
Some sub-processors are located in the USA. We rely on the EU–US Data Privacy Framework certifications (where applicable) and on the EU Standard Contractual Clauses agreed with each processor as the legal basis for transfer. A list of relevant SCCs is available on request from privacy@wovero.com.
6. Local data on your device
The desktop app stores the following locally only — in ~/Library/Application Support/com.wovero.kaho/ on macOS, and ~/.local/share/com.wovero.kaho/ on Linux:
- Encrypted SQLite database with your transcription history and personal correction dictionary
- Cached glossary + profile (re-fetched periodically from our API)
- Application logs
We do not have access to local data unless you explicitly share it with support.
7. Permissions (macOS & Linux)
- Microphone — to capture audio while the hotkey is held
- Accessibility — for global hotkey detection, to read the focused text field for context, and to inject the transcribed text where your cursor is
- Input Monitoring — for the global hotkey listener (CGEventTap on macOS)
On macOS you can revoke any of these at any time in System Settings → Privacy & Security. On Linux, Kaho uses your desktop's microphone permission and (on X11/Wayland) input access for the global hotkey and text injection, which you control through your distribution's settings.
8. Retention
- Audio: 0 seconds — discarded once transcription returns.
- Account data: until you request deletion.
- Usage counters: rolling 13 months for billing reconciliation.
- Logs: 30 days at sub-processors, indefinitely on your local Mac (you can delete them via the app's "Open log directory" button).
9. Your rights (Art. 15–22 GDPR)
You have the right to:
- Access the data we hold about you (Art. 15)
- Have it corrected (Art. 16)
- Have it deleted (Art. 17) — write to privacy@wovero.com and we'll act within 30 days
- Restrict or object to processing (Art. 18, 21)
- Receive a portable copy of your data (Art. 20)
- File a complaint with a supervisory authority — for us that is Der Hessische Beauftragte für Datenschutz und Informationsfreiheit (company seat: Bad Hersfeld, Hessen)
10. Security
All transmissions to our API and sub-processors are TLS-encrypted. The desktop app is signed with an Apple Developer ID certificate and notarized by Apple. Auto-updates are signed with our minisign updater key and verified by the desktop app before installation.
11. Changes to this policy
We may update this policy when we add features or change sub-processors. The "last updated" date at the top reflects the most recent change. For material changes we will notify active users by email.
12. Contact
Questions, requests, or concerns: privacy@wovero.com.