Memorion Privacy Policy

• Email Address — used to create and authenticate your account, send important account-related notifications, and respond to support requests.
• User ID — a unique identifier we generate to associate your data (memories, settings, preferences) with your account. We may also receive a unique identifier from your sign-in provider (Apple ID, Google ID) when you choose Sign in with Apple or Sign in with Google.

• Audio recordings captured through the Memorion wearable device or your phone's microphone, when you actively initiate a recording.
• Transcriptions generated from your audio, including AI-generated summaries, action items, and structured memories.

To produce transcriptions, your audio is sent to one of the third-party speech-to-text providers listed in Section 5.3 (Soniox or Deepgram). To produce AI-generated summaries, action items, structured memories, and chat responses, the resulting transcribed text is sent to one of the third-party AI providers listed in Section 5.4 (OpenAI, Anthropic, or Google Gemini). Each of these providers is contractually prohibited from retaining your data after processing or using it to train their AI models.

Your audio recordings, transcriptions, and memories are stored in our cloud infrastructure (see Section 5.2) so that they are available on every device you sign in to and can be searched by the AI assistant. They are encrypted in transit (TLS) and at rest. End-to-end encryption — where the data would be encrypted with a key only you hold — is planned but not yet available.

• App Diagnostics — the Memorion app does not bundle a crash-reporting or product-analytics SDK, so we do not receive crash reports or usage analytics from your device.
• Server Request Logs — our servers record basic information about each API request — timestamp, endpoint, response time, response status, and your User ID — so we can diagnose errors and monitor performance. These logs do not contain the content of your conversations.

We collect no device-level analytics, and these server logs are never used to profile you or shared with advertisers.

Calendar access is optional and turned off by default. If you enable it, the app reads events from the calendars on your device within a rolling window of the past 7 days and the next 60 days, and syncs the following fields to your account so the AI assistant can answer questions about your schedule:

• Event title.
• Start and end time, including whether the event is all-day and its time zone.
• Location, if you entered one.
• The name of the calendar the event came from, and whether it is a recurring event.

This calendar data is stored in our cloud infrastructure (see Section 5.2) and sent to the AI providers listed in Section 5.4 when needed to answer your questions. Those providers are contractually prohibited from retaining it after processing or using it to train their AI models.

You can turn the calendar toggle off at any time in the app's settings. Doing so permanently deletes all previously synced calendar events from our servers; deleting your account removes them as well.

Your audio recordings, transcriptions, and memories are stored in our cloud infrastructure (see Section 5.2) so that they are available on every device you sign in to and can be searched by the AI assistant. Audio captured by the Memorion wearable is streamed to your phone over Bluetooth and uploaded from there; the phone does not keep a long-term local copy.

Your data is encrypted in transit using industry-standard TLS and encrypted at rest by our storage providers. End-to-end encryption — where your data would be encrypted on your device with a key only you hold, so that not even we could read it — is on our roadmap and is shown as "Planned" in the app. It is not available today, and you should not assume your content is technically inaccessible to us.

We implement appropriate technical and organizational measures to protect your data against unauthorized access, loss, or disclosure. However, no method of electronic storage is 100% secure. We encourage you to use a strong password and enable two-factor authentication on your sign-in provider account.

• Apple (Sign in with Apple) — receives your email (or relay address) and Apple-issued user identifier.
• Google (Sign in with Google) — receives your email and Google-issued user identifier.

• Google Firebase / Firestore (Google LLC) — stores your account profile, conversations, memories, and settings.
• Google Cloud Storage (Google LLC) — stores your audio files and uploaded media in encrypted form.
• Modal (Modal Labs, Inc.) — runs our scheduled background jobs and our speaker-recognition (voiceprint) service.
• Amazon Web Services (Amazon Web Services, Inc.) — hosts our backend API servers (Amazon ECS) and our cache layer (Amazon ElastiCache for Redis), which holds session data and ephemeral state.
• Pinecone (Pinecone Systems, Inc.) — stores vector embeddings of your transcriptions to power semantic search.
• Vercel (Vercel Inc.) — hosts our website at memorion.me and provides aggregated, cookieless traffic analytics. It does not receive your recordings, transcriptions, or memories.

The Memorion app sends your audio recordings to one of the following STT providers for transcription. Audio sent to these providers is not retained by them after processing and is not used to train their models:

• Soniox (Soniox, Inc.) — primary real-time and multilingual transcription provider.
• Deepgram (Deepgram, Inc.) — secondary real-time transcription provider.

The specific provider used depends on the language of your recording and our server-side configuration.

After transcription, your transcribed text may be sent to one of the following LLM providers to generate summaries, structured memories, action items, and chat responses. Data sent to these providers is not used by them to train their models:

• OpenAI (OpenAI, L.L.C.) — accessed via OpenRouter (OpenRouter, Inc.); used for summarization, memory extraction, and chat.
• Anthropic (Anthropic, PBC) — accessed via Amazon Web Services Bedrock (Amazon Web Services, Inc.); used for summarization, memory generation, and chat.
• Google Gemini (Google LLC) — accessed via Google Cloud Vertex AI; used for summarization, memory extraction, text embeddings for semantic search, and image generation for daily reports.

• The Memorion app does not include a crash-reporting or product-analytics SDK (such as Firebase Crashlytics, Sentry, or Mixpanel).
• Diagnostic information is limited to the server-side request logs described in Section 2.3.

We do not sell, rent, or share your personal data with third parties for their own marketing or commercial purposes. None of the providers above are authorized to use your data to train their AI models, build advertising profiles, or pursue any independent commercial use.