Mach records audio only when you start dictation in the Mac app. That audio is sent to Mach's backend and transcription providers so it can be converted into text and optionally polished.

Mach also processes account details, device identifiers, authentication status, app settings, usage counts, diagnostics, and billing-related information when those features are enabled.

Mach stores account and usage metadata needed to run the service, enforce quotas, troubleshoot reliability, and keep your plan in sync across the website and Mac app.

Mach does not store raw audio or transcript text in backend analytics tables by default. Dictation event records are metadata-only, such as timing, status, language, mode, and word count.

Mach uses trusted providers for authentication, hosting, infrastructure, observability, transcription, and language-model processing. These providers process data only as needed to operate Mach.

Current production providers may include Clerk for authentication, Cloudflare for website and network services, Hetzner for backend hosting, Deepgram for speech-to-text, and configured LLM providers for text polishing.

Mach uses HTTPS for network traffic, production authentication through Clerk, server-side quota checks, and signed Mac app releases. Production secrets are kept outside the repository.

No internet service is perfectly secure, but Mach is designed to minimize retained dictation content and keep production access paths narrow.

You can stop using dictation at any time, sign out of the Mac app, or revoke macOS permissions such as Microphone and Accessibility from System Settings.

For account deletion, privacy questions, or data requests, contact hello@usemach.app from the email address connected to your Mach account.

Mach may update this policy as the product changes. Material updates will be reflected on this page with a new effective date.