Privacy & data

What we store, who can see what, and what the AI gets to work with. The legal text follows below – in full.

Read about the methodology

Privacy

Three things that are not up for negotiation.

  1. 01

    Nobody sees anyone else's test answers. That includes parents – they can't see their children's.

  2. 02

    The AI never gets your raw test answers. It works with calculated profiles, identified patterns and the context needed to write the text.

  3. 03

    Nothing is sold, and our AI providers don't train on your data.

How we handle data

Who can see what

Access follows your role in the family and the specific report. If you are not part of a report, you don't see it – regardless of age.

WhatAdult in the familyTeenager (13–17)Child (under 13)The AIPractitioner
Raw test answersOnly your ownOnly their ownNo – the child answers but doesn't see themNoNo
Profile (10 tendencies)Your own – and what others share with youTheir own – they choose to shareThe parent sees the child'sYes, the calculated profileOnly if you share
Individual reportYour own – and children's under 13Only the teenagerNo – adults in the family see itWords the textOnly if you share
Couple reportThe two partnersNoNoWords the textOnly if you share
Parent-child reportThe parent in the reportYes, when the teenager is one of the twoNoWords the textOnly if you share
Sibling reportYes, if a child under 13 is includedYes, if all siblings are 13+NoWords the textOnly if you share
Family overviewThe adults who take partOnly if an adult shares itNoWords the textOnly if you share

The AI receives calculated profiles, the patterns found, and nicknames – never raw test answers or email. Sharing with a practitioner is always initiated by you and can be revoked.

What does SAMRUM store about you?

What we store

  • Account details: email address, language preference and login details.
  • For each person in the family: name, optional nickname, date of birth and role (adult, teenager, child), plus the relationships between family members.
  • What you choose and write in Right now guides and focus tracks, e.g. the situation, free text and check-in answers.
  • Your test answers (only visible to you)
  • Computed profile for report generation
  • Generated reports (text + audio file and transcript, if you chose to generate an audio summary)
  • Purchase history and receipts
  • Consent log with timestamps
  • OAuth tokens for AI assistants you've connected via MCP (stored hashed, auto-expiring, revocable at any time).

How long we keep data

  • Your account, family, profiles, reports, focus tracks and guides are kept for as long as you have an account. You can delete people, families and your account in the app.
  • If you delete your account or a family, access is closed immediately.
  • Our payment provider (Mollie) keeps payment details under its own rules and applicable bookkeeping requirements.

How does consent work in SAMRUM?

Optional consents include product improvement, in-app marketing – and you can sign up for the website newsletter (double opt-in).

  • Product improvement

    Optional

    Allows us to use anonymized data to improve questions and reports. You can opt out at any time.

  • Marketing cookies

    Optional

    Allows Google, Meta, Reddit, and OpenAI (ChatGPT Ads) to measure ad effectiveness. On purchases and signups, a one-way hashed version of your email is sent from our server for improved conversion measurement. Test results are never shared. You can opt out at any time.

  • Website newsletter

    Optional

    If you subscribe without an account, we store your email for double opt-in and emails you asked for. You can unsubscribe in one click from any email or contact us for deletion.

You can use SAMRUM fully without giving consent. You can change it anytime in Settings.

You can adjust your cookie choices at any time here:

How is children's data protected?

Extra rules apply for people under 18.

  • Children under 13 don't have their own login – an adult manages their profile
  • Adults can see children's individual reports
  • Teenagers (13-17) have private individual reports (unless they don't have their own user)
  • Children's results are described carefully – without harsh words
  • Children can never initiate purchases of reports

Who processes data on behalf of SAMRUM?

We use selected service providers to deliver the service. They process data on our behalf and are covered by data processing agreements.

  • Anthropic (Claude)

    Writes reports and the audio summary script. Receives profile results, nicknames, and report content – never raw test answers or email. Anthropic does not train on our data. Covered by the EU-US Data Privacy Framework.

    Report texts and audio summary script

  • Google (Gemini TTS)

    Synthesises the audio summary – only when you actively tap "Generate audio". Receives the script and nicknames. The script is a narration of the report and therefore contains observations about your relationship. The finished audio file is stored with us, not with Google. Covered by the EU-US Data Privacy Framework.

    Speech synthesis for audio summary

  • Resend

    Sends login links, receipts, and notifications.

    Email

  • Mollie

    Handles payment processing. We don't store card details.

    Payments

  • Google (Analytics)

    We use Google Analytics to understand how the site is used – e.g. page views, flows and conversions. Without consent, only anonymous cookieless measurements are collected (no personal data). With analytics consent, full cookie-based measurement is enabled. Data is also sent from our server on purchases and signups. Test results are never shared.

    Analytics

  • Google (Google Ads)

    We use Google Ads to measure ad effectiveness. On purchases, we send a one-way hashed version of your email to Google for improved conversion measurement (Enhanced Conversions). Test results are never shared. Requires marketing consent.

    Advertising

  • Meta (Facebook/Instagram)

    We use Meta Pixel and Conversions API to measure ad effectiveness. On purchases and signups, our server sends a one-way hashed version of your email to Meta for improved conversion measurement. Test results are never shared. Requires marketing consent.

    Advertising

  • Reddit

    We use Reddit Pixel and Conversions API to measure ad effectiveness. On purchases and signups, our server sends a one-way hashed version of your email to Reddit for improved conversion measurement. Test results are never shared. Requires marketing consent.

    Advertising

  • OpenAI (ChatGPT Ads)

    We use the OpenAI Measurement Pixel and Conversions API to measure ad effectiveness for ads in ChatGPT. The pixel stores a first-party attribution ID (oppref) when you arrive from an ad. On purchases and signups, our server sends a one-way hashed version of your email to OpenAI for improved conversion measurement. Test results are never shared. Requires marketing consent.

    Advertising

  • Internal statistics (self-hosted)

    We use self-hosted, cookie-free statistics to understand traffic patterns, bounce rates and page flows. No personal data is collected, no cookies are set and no data is shared with third parties. Does not require consent.

    Traffic analysis

What happens when you connect an AI assistant?

This only applies if you choose to connect your own AI assistant – such as ChatGPT, Claude, or Cursor – to SAMRUM via our MCP server. It never happens automatically. You start the connection from the assistant, grant consent via OAuth 2.1, and the assistant then uses a time-limited token to fetch selected data on your behalf – until you revoke access. All tools are read-only; the assistant can view data, but never change your account, make purchases, or write messages on your behalf. The assistant never sees more than you can see yourself in the SAMRUM app. Here's exactly what the assistant sends us, what each tool sends back, who receives it, how long we keep connection data, and how to disconnect.

Purposes

  • Let you talk to an AI assistant about your family personality data: reports, profiles, focus tracks and Right now guides.
  • Send you to the right places in the SAMRUM app via deep links.
  • Help you find the next step – for example, a focus track that fits a situation.
  • Protect the service against abuse and errors (rate limiting, filtering out repeated calls).

What the assistant sends to SAMRUM

  • A time-limited OAuth access token showing that you have given consent.
  • The tool's input: the language for the answer and – depending on the tool – the ID of the family, person, report or focus track you're asking about, plus the choice of detail level, report section, number of guides and whether to include recommendations.
  • Technical connection details: a session ID and your language setting (e.g. Accept-Language), so we can answer in your language.
  • We never receive your actual conversation with the assistant – only the inputs the tool is called with.

Applies to every answer from SAMRUM

  • People are referred to by nickname – or first name if no nickname is set – and role (adult, teenager, child).
  • Internal, random IDs for families, people, reports, tracks and guides, so the assistant can fetch the next thing. The IDs say nothing about you outside SAMRUM.
  • Links back into the SAMRUM app tagged utm_source=mcp, so we can see the click came from an AI assistant.
  • Fixed instructions to the assistant about tone and boundaries (e.g. describing friction as a loop, never as blame).
  • On errors: an error code, a short explanation and possibly a link – never internal error details.

Tool by tool

This is what goes in and out of each of the 7 tools. The fields match what the server actually returns.

Product information get_samrum_info

Input: Language.

Returns: Fixed product information: what SAMRUM is, methodology, report types, focus tracks, the Right now guide, privacy principles and links. Prices and checkout links only in assistants where that is allowed (not in ChatGPT). No personal data.

Family overview get_family_overview

Input: Language, optional family ID, and whether to include recommendations.

Returns: The family's ID – and family names, but only if you belong to more than one family. For each member: ID, nickname/first name, role, whether it's you, and whether the test is completed. Relationships between members (e.g. partner, parent, sibling). Ready reports (ID, type, participants' nicknames, creation date, link). Active focus tracks (ID, type, name, description, current week, participants' nicknames, link). Up to five latest Right now guides (ID, date, who the guide was about, situation type, recognised pattern, link). Recommended next steps with links.

Latest Right now guides get_recent_moment_guides

Input: Language, optional family ID, number of guides (1–5) and detail level.

Returns: For each guide: ID, status, creation time, whether it was solo or about someone else, the nickname of the person or people the guide was about, situation type, recognised pattern, the type of any referral (e.g. to a report or a professional), an excerpt of the guide's text and – only if requested – the guide's full text. The free text you wrote yourself is not sent as a field, but the guide's text may refer to the situation.

Personality profile and profile chart get_profile · render_profile_radar

Input: Language, person ID and optionally the person ID of one more person to compare with.

Returns: For each person: ID, nickname/first name and role, the profile date, age-appropriate framing for children and teenagers, and the 10 tendencies as coarse bands (high/middle/low) with a short text. A reminder if the test is more than a year old. Short history: number of profiles, the dates of the latest and previous profile and which tendencies have shifted. render_profile_radar shows the same data as a chart in the assistant. Only people whose profile you have access to.

Report get_report

Input: Language, report ID, detail level and optional section.

Returns: The report's ID, type and status. Participants (ID and nickname/first name). A summary of patterns, conflict loops, experiments and conversation starters – or just one section. On request, the report's full text without internal generation data. A flag if the report recommends talking to a professional. Only reports you have access to.

Focus track get_track

Input: Language and track ID.

Returns: The track's ID, type, name, description and status. Participants (ID and nickname/first name). Number of weeks, current week and next delivery date. Linked report ID. Your role in the track and when you last opened it. Check-in status with your own answer, and whether the other participant has answered – not what they answered. This week's content and summaries of previous weeks with delivery date and when you read them.

What's NEVER sent to the AI assistant

  • Full names, email addresses, dates of birth and exact ages.
  • Family names, when you only belong to one family.
  • Raw test answers. They never leave SAMRUM.
  • Precise 5-zone scores, percentiles, or internal axis keys (e.g. "structure_need").
  • Other families you're not a member of.
  • Other family members' raw answers or individual reports they haven't shared with you.
  • When another participant has read a focus track, or what they answered at check-in.
  • Purchase and payment details, card details, passwords, login sessions, or other credentials.
  • Internal generation and debugging data, e.g. AI reasoning, prompts, model details and technical error messages.

Recipients

  • The AI provider you chose – typically OpenAI (ChatGPT), Anthropic (Claude Desktop), or Cursor. They receive the data as an independent data controller and handle it under their own privacy policy, including any transfers outside the EU/EEA.
  • SAMRUM has no control over how the AI provider stores, logs, or uses the data. Read their privacy policy before you connect.
  • We don't share MCP data with advertisers, analytics tools, or any other third parties.

What we store about the connection – and for how long

  • OAuth client registration (the assistant's name, redirect URL, and technical keys) – for as long as the connection exists.
  • Access tokens expire after 15 minutes. Refresh tokens are stored hashed and expire after 30 days – or sooner if you revoke access, change your password or log out on all devices.
  • Operational log on our server: tool name, session ID, duration and error code if any. Contains no names, profile data or response content and is only used for operations and debugging.
  • MCP sessions are held in memory only and cleared after 30 minutes of inactivity. Filtering of repeated calls remembers a call for a few seconds.

Your controls

  • You can revoke an AI assistant's access at any time – from the assistant's own settings (in ChatGPT under apps), or by writing to [email protected].
  • If you change your password or log out on all devices, every connected AI assistant loses access immediately.
  • The assistant only sees what you yourself have access to in SAMRUM. What you don't share in the app, the assistant can't see either.
  • If you delete your SAMRUM account, all access stops immediately.

Technical documentation for the MCP server is available on the developer page.

See the technical MCP documentation

What rights do you have over your data?

Under GDPR, you have a number of rights over your personal data.

  • Access

    You can request access to all the data we have about you.

  • Correction

    You can correct incorrect information.

  • Deletion

    You can delete your account and all associated data – or request deletion by contacting us.

  • Portability

    You can request your data in a structured format.

  • Objection

    You can object to certain types of processing.

  • Restriction

    You can request restriction of processing.

To exercise your rights, you can write to us or use the options in your account.

Who is the data controller?

SAMRUM is the data controller for the processing of your personal data.

Company
SAMRUM, xrplorer ApS
Address
Moesgårdvej 12B, 8270 Højbjerg, Denmark
VAT
DK41348534
Email
[email protected]
Website
https://samrum.io

Questions about privacy?

Write to us if you're unsure about anything.

Write to [email protected]Read about the methodology

Written by Thomas Silkjær, founder of SAMRUMLast updated