Privacy Policy

Our Privacy Commitment

Wisply is built with privacy as a core principle. Core recording, speech transcription, diarization, and supported memo insights run on your device using Apple's on-device frameworks. Optional Screen Understanding on macOS recognizes text locally. In Interview Setup, you can copy recognized text into a question, review it, and choose to send that question to a cloud model. Screen images and video frames are not uploaded by this workflow. Optional features that rely on third-party AI providers (Whisper Coach, Interview Sidekick, Interview Setup, Real-Time Voice AI, and custom AI Text-to-Speech) send audio and/or text off-device to OpenAI and Anthropic, but only when you explicitly enable them and consent in-app. Optional Jev question routing sends interview question or recent transcript text to TypeSafe after you select Jev and grant its separate permission.

This page explains exactly what Wisply captures, where it goes, and how you control it.

Summary

Data Flow: Local Core vs. Optional Cloud AI

The most important thing to understand about Wisply's data handling is the distinction between core on-device features and optional cloud AI features:

• Core features — on-device. When you use Wisply without enabling an optional cloud AI feature, speech, speaker identification, memo insights, and Screen Understanding processing stay on your device. Transcription is performed by Apple's SpeechTranscriber, insights are generated by Apple Foundation Models, and speaker diarization runs through FluidAudio — all locally.
• Cloud AI features — opt-in, data is sent to third parties. Wisply's optional cloud AI features rely on third-party APIs. When you enable one of these features, the relevant data is transmitted directly from your device to the corresponding provider. Specifically:
  • Whisper Coach — sends the audio segment being coached to OpenAI (Whisper / Realtime) for transcription and feedback
  • Interview Sidekick — sends transcript text and chat messages to Anthropic (Claude) and may send transcript context to OpenAI for chat completions
  • Jev question routing — when you select Jev and grant its separate permission, sends interview question or recent transcript text, including available speaker labels, to TypeSafe for classification. Jev selects a question category; Claude generates the answer under its own configuration and Cloud AI consent
  • Interview Setup — sends your reviewed question, selected skill text, and coding preferences to the cloud provider you choose. This includes recognized screen or external-video text you copy into the question. Live transcript context is included only when you enable it; follow-up requests may include the previous generated solution
  • Real-Time Voice AI — streams live microphone audio and transcript context to OpenAI (Realtime API) for live voice-to-voice responses
  • Custom AI Text-to-Speech — sends short text samples (when previewing voices) and playback text to OpenAI (TTS) to synthesize audio

OpenAI and Anthropic features require the in-app "Cloud AI Data Consent" screen. Jev also requires its own TypeSafe permission in Question routing settings. You can revoke these permissions in Settings to block future requests. Stop any active cloud session separately; the Jev revocation behavior is described below.

Information We Collect

Core App Data

Wisply stores the data needed to provide transcription, playback, diarization, and memo history locally on your device by default.

• Audio recordings: If you save a recording, the audio file is stored locally in the app's sandbox and is not uploaded to Wisply servers. Audio segments or live microphone audio can be sent to a third-party AI provider only when you enable a cloud AI feature that requires it — see Third-Party AI Processing below.
• Transcriptions: Stored locally on your device using SwiftData and may optionally sync through your private iCloud account if you enable sync
• AI insights: Where supported, memo insights can be generated on-device by Apple Foundation Models and stored with the memo locally. Optional cloud AI features may generate responses using third-party providers; see below
• Speaker profiles: Stored locally for diarization using FluidAudio
• Context clues: Wisply may read only the frontmost app's name locally to help suggest speaker labels. That lightweight clue is separate from Screen Understanding and does not read the app's visual content
• Screen Understanding: If you enable it on macOS and select a window or display, frames and recognized text are held transiently in memory for local analysis. The capture pipeline does not save or upload them. Text you copy into an Interview Setup question can be sent with that question to your chosen cloud provider after consent

Screen & System Audio Recording Permission

System Audio Capture on macOS

Wisply uses Apple's ScreenCaptureKit framework on macOS to capture system audio that you choose to transcribe, such as audio from a meeting, call, or media application. In System Audio mode, Wisply processes audio sample buffers for transcription and does not analyze visual screen content. The macOS permission is named Screen & System Audio Recording because Apple manages both system audio and screen-content access through ScreenCaptureKit.

Captured system audio is treated like microphone audio: it is processed in memory for transcription and, only if you save the memo, the resulting audio file is written to the app's local sandbox. It is never uploaded to Wisply servers.

Screen Understanding on macOS

Screen Understanding is an optional macOS feature that is off by default. Turning it on does not start capture. For each session, you must use Apple's system content-sharing picker to choose one window or one display. Before capture begins, Wisply requires you to confirm that anyone whose call or content may appear has been informed as needed. This confirmation also applies to calls running in a web browser.

While Screen Understanding is active, Wisply samples the selected content at a limited rate and uses Apple's Vision framework to recognize visible text on your Mac. When Apple Intelligence is available, Apple's on-device Foundation Models framework condenses that text into concise context for Sidekick. If the on-device language model is unavailable, Wisply falls back to recognized text.

Screen frames and recognized text are processed transiently in memory. The capture pipeline does not save them to your Wisply library, write them to logs, sync them through iCloud, or upload them. In Interview Setup, you can copy recognized text into the editable question field. If you choose a cloud model and Generate after granting Cloud AI consent, that question text is sent directly to the selected provider. Screen images and video frames are not uploaded by this workflow. Wisply does not use Screen Understanding content to train a model.

macOS displays its system sharing indicator while capture is active, and Wisply also shows a visible red Screen badge and Stop control. Wisply has no hidden or stealth capture mode. Capture stops when you stop it, turn Screen Understanding off, revoke its permission in Wisply, or close the visible Sidekick surface that started the session.

Screen Understanding is available on macOS. It is not exposed in the iPhone or iPad app.

On-Device Processing and Optional Cloud AI

Speech-to-text transcription, speaker identification, and Screen Understanding run on device by default. Interview Setup also recognizes text from a selected external video input locally. Selecting a screen or video source does not authorize a cloud request. Recognized text enters a cloud request only if you copy it into your question and choose Generate with a cloud model after granting consent.

Optional cloud AI features include Interview Setup, Interview Sidekick, Coding Mode, Jev question routing, Real-Time Voice AI, and custom text-to-speech. Before those features send transcript text, prompts, selected synthesis text, or live microphone audio to a named provider, Wisply presents a consent screen describing the data and recipient. Those requests go directly from your device to the provider; Wisply does not operate a proxy server.

Interview Setup Requests

Interview Setup lets you choose Claude from Anthropic, OpenAI, or the existing on-device coding assist. When you choose a cloud provider and generate an answer, Wisply sends your reviewed question, selected skill text, and coding preferences directly to that provider. Live transcript context is included only when you turn on "Include Wisply's live transcript". Follow-up requests may also include the previous generated solution.

Cloud requests require in-app Cloud AI consent and use your provider API key, stored in Keychain. Saving a key, opening Interview Setup, or reviewing a tutorial does not send a model request. Provider billing and model access are checked when a request runs. You can remove a provider key or revoke Cloud AI consent in Settings. The OpenAI key is shared with Real-Time Voice AI, so removing it affects both features.

Your Controls

You decide whether to enable Screen Understanding and what single window or display to share for each session. You can stop an active session at any time from the visible Sidekick controls, turn the feature off in Settings, or revoke Screen & System Audio Recording access in macOS System Settings. Screen Understanding is not active when its visible Wisply surface is closed.

Third-Party AI Processing

Wisply's optional cloud AI features rely on third-party AI providers. These calls happen directly from your device to the provider — Wisply does not operate a server that proxies, logs, or stores the requests or responses. Content is sent only after you enable the relevant feature and grant its required in-app permission. OpenAI and Anthropic features use "Cloud AI Data Consent"; Jev question routing requires a separate TypeSafe permission. Interview Setup questions can include recognized screen or external-video text that you copy into the question and choose to send. Screen images and video frames are not uploaded by that workflow.

• OpenAI — used for Whisper Coach (audio transcription and feedback), Real-Time Voice AI (streaming audio + transcript context through the Realtime API), Interview Setup and chat-based Interview Sidekick completions, and custom AI Text-to-Speech (short text samples when previewing voices and playback text when synthesizing speech). See OpenAI's privacy policy.
• Anthropic — used for Interview Setup and Interview Sidekick requests to Claude, including the question, prompts, and context described above. See Anthropic's privacy policy.
• TypeSafe — used for optional Jev classification of interview questions and recent transcript text, including available speaker labels. Jev does not generate the interview answer. See TypeSafe's privacy policy.

What gets sent depends on the feature you enable:

• Whisper Coach: the audio segment being coached
• Interview Sidekick: your chat messages and selected transcript context
• Jev question routing: the relevant interview question or recent transcript text, including available speaker labels; no microphone audio, screen images, video frames, or full recording archive
• Interview Setup with a cloud model: your reviewed question (including recognized text you copy into it), selected skill text, coding preferences, optionally enabled transcript context, and any previous solution included in a follow-up
• Real-Time Voice AI: live microphone audio and ongoing transcript context
• Custom AI Text-to-Speech: short preview text and the text you ask Wisply to speak

You can revoke Cloud AI consent at any time in Settings to block future OpenAI and Anthropic requests. Stop any active cloud session separately. TypeSafe permission can be revoked in Question routing settings as described below.

OpenAI and Anthropic may retain API inputs and outputs for abuse and safety monitoring for up to 30 days by default. Retention follows the terms and data controls of the provider API account; shorter or zero-data-retention terms apply only when the provider organization has them enabled.

Optional Jev Question Routing

You can choose Jev from TypeSafe to classify interview questions and route them to the appropriate answer flow. This requires a TypeSafe API key and a separate Jev permission in Question routing settings. Saving a key or accepting the general Cloud AI consent screen does not enable Jev.

When Jev is selected, its permission is granted, and Interview mode is active, Wisply sends the relevant question or recent transcript text, including available speaker labels, directly from your device to TypeSafe at api.typesafe.ai. This request does not include microphone audio, screen images, video frames, or your full recording archive. Jev selects a question category; Claude generates the response using its separately configured key and Cloud AI consent. Jev does not change the model used for recording chat or summaries.

Your TypeSafe key is stored in your device Keychain. You can remove the key, choose a different question router, or revoke Jev permission in Question routing settings. Wisply cancels active routing work and its pending answer and discards late results after permission is revoked. Text already transmitted cannot be recalled. Data already received by TypeSafe is governed by your TypeSafe account terms and its privacy policy linked above. Wisply does not promise zero retention for ordinary TypeSafe accounts; the OpenAI and Anthropic retention statement above does not apply to TypeSafe.

Subscription Management (RevenueCat)

Where subscription features are offered, Wisply uses RevenueCat to manage purchases through the App Store:

• Purchase Metadata: Product identifiers, entitlement status, renewal status, and purchase/expiration dates used to verify subscription access
• RevenueCat App User ID / Device Identifier: Used for subscription management across devices (not used for advertising or cross-app tracking)
• Purpose: App functionality only - managing purchase entitlements

Waitlist

If you join our waitlist, we collect your email address to notify you when the app launches. We do not share this with third parties or use it for other purposes.

Optional Wisply Account and Sign-In

You can record and transcribe without creating a Wisply account. If you choose Sign in with Apple or Google, the app sends the provider's identity token or authorization code to Wisply's authentication service so we can verify the sign-in and create or retrieve your account.

• Account identifiers: A Wisply user ID and the Apple or Google provider user ID
• Contact and profile data: Email address, display name, and—when supplied by Google—profile image URL
• Account activity: Account creation and last sign-in timestamps, linked provider, and subscription status
• Purpose: Authentication, account management, support, and optional account features—not advertising or cross-app tracking

Wisply stores this account data in its protected backend database. Authentication credentials issued to the app are stored in the device Keychain. You can delete your Wisply account from Settings > Account > Delete My Account; this removes the account and linked server-side account records. Local recordings and transcripts remain on your device until you delete them in Wisply or remove the app.

Subscriptions and Cloud Sync

In releases where subscriptions or iCloud sync are available, purchases use Apple App Store in-app purchase and sync remains optional. If you enable iCloud sync:

• Transcript and memo data may sync through Apple's private CloudKit database to your devices
• Audio recordings remain local to the device and are not synced through CloudKit
• Wisply does not sync this content through Wisply servers
• Subscription status is managed through Apple's App Store and RevenueCat

On-Device Technologies

Wisply uses the following Apple technologies for core and other non-cloud features. These process data locally on your device and do not transmit it externally:

• Apple Speech Framework (SpeechTranscriber): Real-time, on-device speech-to-text transcription
• Apple Foundation Models: On-device AI for generating supported summaries, insights, and concise Screen Understanding context on compatible devices
• Apple Vision: On-device recognition of visible text from the window or display you select for Screen Understanding
• SwiftData: Local database for storing your memos and transcripts
• ScreenCaptureKit: System audio capture and, only when you opt in and make a per-session selection, transient capture of one window or display for Screen Understanding
• Frontmost App Name: Used locally as a lightweight context clue for speaker labeling without reading the app's visual content
• File Timestamps: Used for sorting and displaying your memos chronologically
• Disk Space: Checked to manage audio file storage
• UserDefaults: Stores app preferences and settings locally

When you enable an optional cloud AI feature, the transcription or reasoning step for that specific feature is handled by the third-party provider described in the Third-Party AI Processing section above, rather than by the on-device frameworks listed here.

Third-Party: FluidAudio

Wisply uses FluidAudio for speaker diarization (identifying who said what). FluidAudio processes audio locally on your device and does not transmit any data externally.

Third-Party: RevenueCat

For subscription management, we use RevenueCat which accesses system boot time for device identification purposes. This is used solely for managing your subscription across devices and preventing unauthorized access.

Permissions

Wisply requests only the permissions needed for features you use. On macOS these include:

• Microphone: Required to transcribe your voice
• Screen & System Audio Recording: Required by ScreenCaptureKit for optional system audio capture and optional Screen Understanding. Screen Understanding remains off by default and captures only the window or display you select for that session
• Camera: Required only when you choose an external video input for local text recognition in Interview Setup
• Accessibility: Optional for automatic typing into another app in builds that support it. Global keyboard shortcuts do not require Accessibility access

On iPhone and iPad, Wisply requests Microphone and Speech Recognition for recording and transcription; notifications are optional. On Apple Watch, microphone access is used for watch recordings, which are transferred to the paired iPhone for transcription.

You can revoke these permissions at any time in System Settings > Privacy & Security.

Data Storage

Wisply stores your data locally on your device by default:

• Audio recordings are stored locally in the app's sandbox as files associated with your memo
• Memos and transcripts are stored in a SwiftData database locally by default
• Where available and enabled, transcript and memo data may sync through your private iCloud account
• Speaker embeddings are stored locally for consistent speaker identification
• App preferences are stored in UserDefaults
• If you create an optional account, account identifiers, email, display name, provider information, and account timestamps are stored on Wisply's protected backend
• Screen Understanding processes frames and recognized text transiently in memory without saving them to local storage, logs, or iCloud. If you copy recognized text into an Interview Setup question and send it to a cloud model, the provider processes that text under its own retention terms

To delete local data, you can delete individual memos within the app or remove the app and its local container. If iCloud sync is enabled, delete synced memos from within the app so the change propagates to your private iCloud database.

Data Sharing

We do not sell your data. We do not share your personal information with third parties for marketing purposes.

We do not share screen content, audio content, or transcript content with third parties for advertising or marketing. We may disclose limited information only in the following circumstances:

• With Apple and RevenueCat for App Store purchase and subscription management
• With Apple or Google when you explicitly use that provider to sign in
• With Wisply's database and hosting providers acting as service providers for optional account authentication and storage
• With Apple's private CloudKit database, but only if you enable iCloud sync for transcript and memo data
• With OpenAI and/or Anthropic only if you explicitly enable an optional cloud AI feature (Whisper Coach, Interview Sidekick, Interview Setup with a cloud model, Real-Time Voice AI, or custom AI Text-to-Speech) and grant in-app consent
• With TypeSafe only if you select Jev question routing and grant its separate permission in Question routing settings
• With TelemetryDeck for anonymous product interaction, performance, error, and diagnostic analytics that exclude audio recordings, transcript content, and screen content
• With your explicit consent
• To comply with legal obligations
• To protect our rights or safety

Analytics (TelemetryDeck)

We use TelemetryDeck for privacy-friendly analytics to understand how the app is used and improve it:

• Product Interaction: Anonymous usage statistics (e.g., which features are used, app sessions)
• Performance Data: Anonymous timing, stage, value, and unit measurements used to improve reliability
• Other Diagnostics: Anonymous error type, domain, and code metadata without transcript or audio content
• Not linked to your identity: Data is anonymized and cannot be tied back to you
• Not used for tracking: We do not track you across apps or websites
• No content data: Your audio recordings, transcript content, and screen content are never sent to TelemetryDeck
• Purpose: Analytics only - to improve app functionality and user experience

TelemetryDeck is privacy-focused and GDPR-compliant. No advertising IDs or personal identifiers are collected.

Children's Privacy

Wisply is not directed at children under 13. We do not knowingly collect information from children under 13.

Your Rights

Depending on your location, you may have rights under data protection laws including:

• Access: Request a copy of your data
• Deletion: Request deletion of your data
• Portability: Export your data in a standard format
• Correction: Correct inaccurate data

Since recordings, transcripts, and memo data are stored locally by default, you have significant control over them. If you enable iCloud sync, your synced transcript and memo data is stored in your private iCloud account. To exercise these rights for any limited data we may hold directly (for example, optional account information or a waitlist email address), use Delete My Account in the app or contact us at privacy@wisply.ai.

California Privacy Rights (CCPA)

California residents have additional rights under the California Consumer Privacy Act:

• Right to know what personal information we collect
• Right to delete personal information
• Right to opt-out of the sale of personal information (we do not sell data)
• Right to non-discrimination for exercising your rights

European Privacy Rights (GDPR)

If you are in the European Economic Area, you have additional rights under GDPR:

• Right to access your personal data
• Right to rectification of inaccurate data
• Right to erasure ("right to be forgotten")
• Right to data portability
• Right to object to processing
• Right to lodge a complaint with a supervisory authority

Changes to This Policy

We may update this Privacy Policy from time to time. We will notify you of any changes by posting the new Privacy Policy on this page and updating the "Last updated" date.

Contact Us

If you have any questions about this Privacy Policy, please contact us:

• Email: privacy@wisply.ai
• Support: wisply.ai/support