AI and your data
Last updated: 2026-10-18
MemoSeek uses AI services, including OpenAI, to understand your requests. Information needed to process a request goes to the selected provider. ElevenLabs may receive microphone audio after you press Send if its transcription route is selected and you authorize that purpose, and text for generated spoken replies. Playing bundled prerecorded voices sends it neither your words nor microphone audio. You can review the providers and purposes and change your choice in Privacy settings.
What uses AI
- Understanding what you say or type, to turn it into a reminder or a saved detail.
- Reading the photos you add and pulling out their details, such as dates and numbers. For an identity document these details can include the document type, holder name, document number, nationality, issuing country, date of birth, and issue and expiry dates.
- Generating spoken replies that are not pre-recorded.
Your permission
The app asks for your permission before first use. You can change your choice at any time in Settings → Privacy & Data → AI processing. If you choose "Not now", MemoSeek works in a limited mode: you can browse, edit and delete what you have saved, but voice, typed requests and photo reading are off.
| Provider | Used for | May receive |
|---|---|---|
| Google (Gemini API) | Understanding requests, reading selected photos/documents, speech recognition and voice conversation where enabled, and speech generation. | Your request, relevant saved or conversation context, selected images/documents, microphone audio in enabled voice modes, and text for spoken replies. |
| OpenAI (API) | Understanding requests, reading selected photos/documents, speech-to-text where enabled, and speech generation. | Your request and relevant context, selected images/documents, microphone audio when its transcription or voice adapter is enabled, and text for spoken replies. |
| Anthropic (Claude API) | Understanding and analyzing requests. | Request text and relevant conversation or saved-item context. No microphone audio is sent to this text adapter. |
| Deepgram | Turning speech into text when the Deepgram transcription route is selected. It is not used for spoken replies. | Microphone audio for transcription and language settings. The resulting text is then handled by the selected understanding provider. |
| AssemblyAI | Turning speech into text when the AssemblyAI transcription route is selected. It is not used for spoken replies. | Microphone audio for transcription and language settings. The resulting text is then handled by the selected understanding provider. |
| ElevenLabs | Speech transcription after you press Send when this route is selected and you have authorized that purpose, and generation of spoken replies. Playing bundled generic recordings does not call ElevenLabs. | Microphone audio submitted after Send for authorized transcription, and text for generated spoken replies. Playing bundled clips sends no user requests or microphone audio. |
Your voice
- MemoSeek does not store raw recordings of your voice.
- In the Apple speech mode, Apple's speech recognition on your iPhone turns your speech into text, and Apple may process that speech on its servers under Apple's terms. That processing is between you and Apple.
- ElevenLabs can receive audio after you press Send when its transcription route is selected and you have authorized that purpose. Audio can also reach Deepgram or AssemblyAI for speech-to-text when that route is selected; neither generates spoken replies. Your audio reaches AssemblyAI only if you agreed to the provider list that names it.
- In some voice modes, your raw voice audio is streamed to Google's Gemini API (or to OpenAI, if its voice mode is enabled for your account) so that it can be understood directly.
Spoken replies
Generic spoken phrases, like a short acknowledgement, are pre-recorded and shared by everyone. Your personal content never goes into this shared voice library.
Providers' terms
We use these providers' paid API services. Their terms govern how they handle data; we do not claim more than their published terms say:
AI can make mistakes
AI can get a time, a name or a number wrong. Review important dates and details yourself. The full detail is in the privacy policy and on Subprocessors.