Privacy Policy
Last updated: November 8, 2026
1. Core Philosophy: Zero-Retention Voice Privacy
At Sawt, we believe your spoken thoughts and words are private. Unlike web-based dictation platforms that retain your audio for model training, Sawt is designed with a strict zero-retention architecture. Your audio streams are processed in real-time solely to return text to your Windows cursor.
2. Local Storage & History
All your transcription history, preferences, and custom vocabulary dictionary are stored locally on your machine in an encrypted SQLite database. Sawt does not operate central cloud databases that aggregate user keystrokes or dictation histories.
- Location: Stored under your local Windows AppData directory.
- Control: You can view, search, export, or completely purge your history with one click inside the Sawt desktop settings.
3. Audio Transmission & Encryption
When you press Ctrl + Shift or Ctrl + Alt, your microphone records in memory. The audio buffer is securely transmitted via TLS 1.3 encryption directly to the designated AI speech engine (e.g. OpenAI Whisper or Groq cloud endpoint).
Neither Sawt nor our transcription infrastructure logs or retains audio snippets after transcription finishes. Audio bytes exist only in RAM during the active dictation call and are discarded immediately upon completion.
4. Third-Party API Keys (BYOK)
If you configure your own API key (Bring Your Own Key mode), all network requests go directly from your PC to the API provider. Your API key is encrypted using Windows DPAPI (Data Protection API) and never touches our servers.
5. Telemetry & Analytics
Sawt collects zero identifiable telemetry, zero keystrokes, and zero voice prints. We do not use third-party analytics trackers (such as Google Analytics or Facebook Pixel) inside the desktop application.
6. Contact
If you have questions regarding this policy or data safety, please reach out to us at support@sawt.ai.