Technical Transparency

How Your Data Is Handled

A plain-language breakdown of what happens to your transcripts, and when.

Section 1 - What you upload

Meeting transcripts are saved to your project in Supabase, a secure hosted database. The transcript is stored as entered by you so your meeting history remains complete in your account.

Section 2 - What the AI sees

Privacy ModeWhat AI Receives
Fully OpenOriginal transcript
Semi-PrivateTranscript with emails, phones, URLs, addresses, and names filtered out
Fully PrivateHeavily redacted transcript - most identifying terms are replaced before processing
Zero AINo transcript content is sent to an external AI provider

The server applies this filter before an AI request is made. The provider processes only what your selected mode allows through.

Section 3 - Who processes it

When AI is enabled, Datum Notes may use external model providers including Google Gemini and Anthropic depending on the workflow. Privacy rules are applied server-side before transcript content is sent to an AI provider.

Provider policies: Google policies.google.com/privacy and Anthropic anthropic.com/privacy.

Section 4 - Google Calendar (optional)

If you connect Google Calendar, Datum Notes requests read access to your primary calendar and permission to read event details (title, time, attendees, meeting links). We store OAuth access and refresh tokens in your account so we can sync upcoming events. Calendar data is used to show your schedule in the Events workspace and to link meeting transcripts to calendar entries. We do not modify or delete calendar events unless you explicitly use a write-capable feature in the future.

You can disconnect by revoking Datum Notes in your Google Account security settings or signing out. Tokens are removed from our database when you delete your account.

Section 5 - What we log

The privacy mode applied to each parse is recorded for auditability. Transcript content is not included in logs.

Section 6 - Your control

You set privacy mode per project. The server resolves and enforces this setting on every AI-enabled request. If no privacy mode is configured, the system defaults to Fully Private as a conservative fallback. Zero AI keeps project data out of external AI workflows.