Privacy Policy

Last updated: 2026-10-25. This policy is a working draft written from how the product actually behaves. Have it reviewed by a lawyer before you rely on it.

ReMind turns meetings into a persistent, searchable record of decisions, commitments and open questions. This page says plainly what happens to your data. Questions: use the contact in the app footer, or the address in our security contact for security matters.

Who is responsible

For the personal data of people who use ReMind (account holders), the operator of this ReMind instance is the controller. For the content of meetings that a customer analyses on behalf of their organisation, the customer is the controller and the operator is the processor; a data processing agreement (DPA) is available on request.

What we receive

A meeting reaches us in one of these ways, and we only receive what that path carries:

What we do with a meeting

  1. The text is held in memory while it is analysed. Audio is transcribed and dropped. When cloud transcription is on, recorded audio is sent to AssemblyAI to be turned into text, and ReMind deletes the audio and the transcript from AssemblyAI as soon as the text comes back. On the Free plan a recording is limited to 30 minutes and 3 recordings a week.
  2. To analyse it, the transcript is sent to language-model providers (see the sub-processors). For each meeting we may send it to up to three independent models so their answers can be cross-checked. The models are run by Groq, Google (Gemini API) and NVIDIA (build.nvidia.com). Google and NVIDIA are used on their free developer tiers, whose terms may allow them to log prompts and use them to improve their products; Groq's terms say it does not train on API inputs. Where you or your organisation operate the instance, you choose the providers, and you can switch the Google and NVIDIA ones off. OpenRouter-style ":free" models stay blocked for customer data in production.
  3. A numeric summary (an embedding) is computed on our own server.
  4. We then keep only the structured result: decisions, commitments, questions, blockers, the short verbatim quotes that are evidence for each item, a mind map, a title and a date. Audio is never stored. The transcript and notes are kept with the meeting, encrypted (AES-256-GCM), only while you leave "Keep transcripts and notes" on in Settings, Your data. It is on by default for new analyses, you can turn it off at any time, and you can delete everything already kept with one button. They are shown only to people who can open that meeting, and are deleted with the meeting or the account. Meetings analysed before this existed have none. One more case: if the AI service is at capacity when your meeting is analysed, the meeting is still analysed and emailed straight away, and the transcript is held encrypted for up to 48 hours only so the project memory (decisions timeline, what changed) can be added later. It is deleted the moment that is done.

We do not use your data to train models, and we do not sell it. The one caveat is the free-tier terms of the third-party model providers listed above, which we do not control.

What we store, and where

DataPurposeRetention
Account email, sign-in identityAccount and email deliveryUntil you delete the account
Meetings: notes, quotes, mind map, title, dateThe productUntil you delete them, or the operator's retention limit (DATA_RETENTION_DAYS)
Transcript and notes of a meeting (encrypted; only if "Keep transcripts and notes" is on)Showing the exact words behind a decision or ticketUntil you delete them in Settings, or the meeting or account is deleted
Project memory (decisions, questions over time)The productDeleted with the project or account
Commitments, assignments, correctionsFollow-through, qualityUntil deleted
Integration tokens (Google, Slack, Notion, GitHub, Linear)Features you switched onEncrypted (AES-256-GCM); removed when you disconnect or delete the account
API keysBrowser extensionStored only as a hash
Transcript waiting for project memory (only when the AI service was at capacity)Adding project memory a little laterEncrypted; deleted as soon as the memory is built, and in any case within 48 hours
Processing job recordsShowing analysis progress and review30 days
Email delivery eventsPrevent duplicate emails, measure delivery400 days
LLM usage records (model, token counts)Capacity and cost400 days
Security audit log (who did what, IP address)Security and accountability400 days

Data is stored in Supabase (PostgreSQL) with row-level security so a person can only read their own rows and, in a shared workspace, what the workspace allows. Data is encrypted in transit (TLS) and at rest by the hosting providers.

Your rights (GDPR articles 15 to 22)

Lawful bases

Contract (providing the service you asked for), consent (integrations, recording, the weekly digest) and legitimate interests (security, preventing abuse, measuring delivery). Recording a call involves other people: tell participants and, where the law requires it, obtain their consent.

International transfers

Some sub-processors are in the United States. Where personal data leaves the EEA we rely on the provider's Standard Contractual Clauses or an adequacy mechanism. The current list and locations are on the sub-processors page.

Security

See /security. Breaches that affect personal data are notified to the authority within 72 hours and to affected people without undue delay.

Cookies and tracking

We set no advertising or analytics cookies and load no third-party fonts or scripts on our own pages. The sign-in session is kept in your browser's local storage. Emails contain no tracking pixel; the button in a change email goes through a signed link that records one click event so we can tell whether the email was useful.

Changes

We will update this page and its date when behaviour changes.


ReMind