Privacy Policy

Last updated: September 18, 2026

ThinkScribe has no accounts and no server that receives your content. We do not collect, store, or have access to your audio or transcripts. Transcription runs on your device.

1. Overview

ThinkScribe ("the App") is a native application for Mac (Apple Silicon), iPhone, iPad, and Apple Watch. The App provides speech-to-text transcription, AI-powered summarization, and meeting intelligence features. Speech-to-text transcription runs on your device using Apple's Neural Engine and local machine learning models (such as WhisperKit, Parakeet TDT, and Apple's on-device speech recognition). AI features use either ThinkScribe AI (powered by Qwen), which runs on your device, or services that Apple provides as part of the operating system (see Section 5).

This Privacy Policy describes what data the App accesses, how it is used, and your rights regarding that data.

2. Data We Do NOT Collect

We, the developer of ThinkScribe, do not collect, receive, or store any of the following:

Audio recordings or microphone input
Transcription text or AI-generated summaries
Personal identification information (names, emails, phone numbers)
Usage analytics, behavioral profiling, or advertising data
Advertising identifiers (IDFA/IDFV), your Apple ID, or your IP address
Location data

3. Data Stored on Your Device

The following data is created and stored locally on your device, within the App's sandboxed container or your chosen export locations:

Transcripts: Text generated from your audio, stored via SwiftData in the App's local database.
Audio files: Recordings captured during sessions, stored in the App's documents directory.
AI model caches: Whisper and LLM model weights downloaded for on-device inference.
User preferences: Settings and configuration stored in UserDefaults.
Search indexes: Local indexes built for fast transcript search.

You have full control over this data. You can delete individual transcripts from within the App, or remove all data by deleting the App.

Optional iCloud Sync: ThinkScribe offers an optional "Sync Across Devices" feature that is turned OFF by default. When you choose to enable it, your transcripts, summaries, collections, and speaker profiles are synced across your own Apple devices using Apple's first-party CloudKit service and your personal iCloud account. This data is stored in your private iCloud database and protected by Apple's iCloud security; we, as the developer, have no access to it. Audio recordings are synced only if you also turn on "Sync Audio to iCloud". No data is sent to us; CloudKit communicates only with Apple's iCloud servers, and you can disable sync at any time in Settings.

4. Permissions

ThinkScribe requests the following permissions:

Microphone: To capture audio for transcription. Audio is transcribed on your device and is never sent to us.
System Audio Recording (Mac only): To capture system audio from applications (e.g., Zoom, Teams). No screen content is involved, only audio streams.
Accessibility (Mac only): To enable global keyboard shortcuts for quick recording access.

5. Third-Party Services

ThinkScribe does not use any third-party analytics, advertising, or tracking services, and contains no SDKs that transmit your content to us or to third parties. Some features use services that Apple provides as part of the operating system, such as Apple Intelligence, Apple's Translation, Apple's speech recognition, and iCloud. Apple decides how these are processed, which may include Apple's servers (for example Private Cloud Compute), under Apple's Privacy Policy. We never receive this content. If a future version offers a feature that sends your content to any other service, that feature will be optional, off by default, and will ask for your permission first.

Aside from the optional iCloud Sync, the Apple-provided services and the anonymous crash diagnostics described in this section, the App's outbound network activity is limited to downloading open-source machine-learning model weights (such as Whisper, Parakeet, and Qwen) from public model repositories on Hugging Face during initial setup. These are one-way downloads of static model files, no audio, text, transcripts, or any user content is uploaded to these services as part of this process. Hugging Face acts as a static file host (functionally equivalent to a public release asset on GitHub), not as an AI service. After models are cached locally, recording and transcription work without an internet connection.

Crash & performance diagnostics: To find and fix bugs, crashes, and freezes, ThinkScribe can send anonymous diagnostic reports to our own servers (hosted by our infrastructure provider, Vercel). Each report contains only a random per-install identifier (regenerated if you reinstall, it is not your Apple ID, name, email, IDFA, or IDFV), the app and operating-system version, your device model, the technical exception or signal type, and the crash's call stack. It never includes your audio, transcripts, summaries, or any personal content, and we do not store your IP address. This data is used solely to diagnose stability problems; it is never sold, never used for advertising or tracking, and never shared with any third party. It is on by default, you can turn it off at any time in Settings ▸ Privacy ("Share Anonymous Diagnostics and Usage"). ThinkScribe uses no third-party analytics or crash-reporting SDKs; the only other diagnostics are the anonymized, aggregated metrics Apple provides to all developers through App Store Connect (via MetricKit), which remain within Apple's systems and cannot identify you.

Anonymous usage statistics: The same report can also carry simple counts of how the app is used: how often a feature, the paywall or the review prompt came up, which onboarding steps were reached, how a purchase attempt ended, a few on/off settings (for example whether iCloud Sync is on, which colour theme and which plan), and one-time milestones such as the first recording. These are counts with fixed names, tied only to the same random per-install identifier. They never contain anything you recorded, imported, typed or named (no titles, speaker names, vocabulary, search text or prompts), no contact details, and no location beyond the country our host derives from the request (the IP address itself is not stored). We use them to understand which features are used and whether the app works as intended. They are never sold, never used for advertising or tracking, and never combined with data from other companies. The same switch turns them off: Settings ▸ Privacy ▸ "Share Anonymous Diagnostics and Usage". Organizations can also disable it by device-management policy.

6. In-App Purchases

ThinkScribe uses Apple's StoreKit 2 framework for in-app purchases. Purchase transactions are handled entirely by Apple and are subject to Apple's Privacy Policy. We do not have access to your payment information.

7. Children's Privacy

ThinkScribe is not directed at children under 13. We do not knowingly collect any personal data from children, as we do not collect personal data from any users.

8. Changes to This Policy

We may update this Privacy Policy from time to time. Changes will be reflected on this page with an updated "Last updated" date, and material changes will also be announced in the App's release notes. As ThinkScribe has no user accounts, we cannot notify you by email.

9. Contact

If you have any questions about this Privacy Policy, please contact us at support@thinkscribe.ai