Privacy policy
Local-first means your meeting stays yours.
Who this policy covers
This policy covers the Recognizing app for Mac and Recognizing Captions for iPhone and iPad. Recognizing does not require a product account and does not operate an advertising, analytics or developer server that receives your meetings.
Recording and consent
Recording starts only after an explicit action and remains visibly indicated. You are responsible for informing participants and obtaining any consent required where you record. Recording laws vary by location and circumstances.
Mac: local processing and storage
Recognizing can record microphone audio and, only when you enable it, system or meeting-app audio. Apple Speech transcribes the recording locally. A meeting bundle may contain its title, transcript, notes, raw audio, screenshots you explicitly capture, screenshot captions, context and OCR text, generated records and privacy-safe diagnostics.
Provider settings are stored locally. Cloud-provider API keys are stored in the macOS Keychain.
Mac: optional cloud processing
Automatic, Apple Intelligence, local MLX and deterministic modes keep meeting content on your Mac. If you choose Cloud (BYOK), Recognizing sends bounded transcript, note, screenshot-caption, context and OCR text directly to the endpoint you configure to generate a record or answer a question. Raw audio and screenshot pixels are not sent. The selected provider may retain or use submitted text under its own terms and privacy policy.
iPhone and iPad
Recognizing Captions records microphone audio only after you explicitly start a session. Version 1.0 records in the foreground. In version 1.0.1, you can enable Continue when locked for an individual conversation; this keeps microphone recording active when you lock the screen or switch apps. The option is off by default for the next conversation. Apple Speech transcribes the conversation on the device. The app does not upload conversation audio or text. A saved conversation may contain its title, transcript, timestamped notes, highlighted moments, photos you explicitly attach, on-device OCR text, a generated local record and raw microphone audio.
Export uses Apple's system share sheet. A Recognizing mobile capture can contain captions, notes, user-attached photos, on-device OCR text and a generated record, but excludes raw audio, local file paths and credentials. You choose the destination, and the receiving service's privacy terms apply after you share.
Continue on Mac
The Mac app imports a mobile capture only after you choose it with the system file picker. The imported recording and attached photos remain on the Mac. If you later choose Cloud (BYOK) analysis, the Mac app shows its off-device disclosure before sending bounded text to the endpoint you configure. Photo pixels are not sent.
From Mac 1.0.2 and iOS 1.0.1, you can explicitly export a Mac record and receive it on the original iPhone or iPad conversation. The returned copy contains exported text, citations and any review warning, without raw audio, image pixels or credentials. You preview and accept it before it is stored separately; the original conversation remains unchanged. This is file sharing, not automatic cloud sync.
Optional on-device image descriptions
From Mac 1.0.2 and iOS 1.0.1, Describe Image processes a saved image only when you choose that action. It requires macOS or iOS 27 and a ready, compatible Apple Intelligence model. The image and request stay on the device. There is no Private Cloud Compute or automatic cloud fallback for this feature. Descriptions are shown separately for you to check against the image and are not automatically added to meeting evidence.
Permissions
Microphone, Speech Recognition and Camera permissions are requested only for features that need them. Choosing an existing photo uses Apple's system picker rather than broad photo-library access. On Mac, Screen Recording is requested only for optional system audio or screenshots. On iPhone and iPad, recording pauses when the app enters the background or the device locks unless you enabled Continue when locked for that conversation in a version that supports it. Audio interruptions and microphone-route changes still pause recording. Recording never resumes automatically after an interruption.
Retention, export and deletion
Whole meetings are kept until you delete them. Raw-audio and diagnostic retention can be changed in Settings where available. You can export supported records, delete an individual recording or use Delete All Local Data to remove Recognizing-owned local data and settings. On Mac, complete deletion also removes Recognizing cloud-provider credentials from Keychain. Data already sent to a provider must be managed with that provider.
Tracking and analytics
The apps do not include advertising, cross-app tracking or product-usage analytics SDKs. Recognizing does not sell meeting data.
Changes and questions
Material changes will be reflected on this page with a new updated date. For privacy questions, use the working contact route on the Recognizing support page.