Privacy
A draft for discussion before launch. Effective date: [EFFECTIVE DATE].
Before publication
[CLARIFY BEFORE PUBLICATION] This draft describes the server code implemented so far, rather than verified operation of every app feature. The mobile app currently calls Gemini directly and uses formsubmit; both flows must move to the server before launch. Production has not been deployed; mobile app connectivity, R2, purchases and external notifications still need verification. Operator: [OPERATOR: legal entity name and address]. Contact: [SUPPORT E-MAIL].
1. Who MOOM is for and why data is needed
MOOM helps parents keep family records, a calendar and AI conversations. Information about children is entered by a parent or an authorised family member. The app is not intended for children to use independently. [AGE RESTRICTIONS AND REPRESENTATIVE AUTHORITY] must be agreed before launch.
A lawyer must determine [LEGAL BASES FOR PROCESSING] and the applicability of the GDPR, Georgian personal data protection law and Russian Federal Law No. 152-FZ, taking account of the operator, markets and data categories. This draft does not establish the applicable law.
2. What data is processed
- Account: name, email (if provided), Apple/Google sign-in identifier, language and time zone settings, and session details. Original ID tokens are verified but not stored; refresh tokens are stored as hashes.
- Family, members, roles and permissions by section; child and pregnancy profiles, dates, allergies, dietary restrictions and entered notes.
- Health, sleep, nutrition, development, pregnancy and emotional support records, recipes, events and reminders.
- Photos and their metadata, AI conversations and messages, proposed actions, results and AI usage records.
- Device push tokens, platform, app version, language and time zone; subscription and trial status, and minimal payment event receipts.
- Consent history, feedback text and voluntarily provided name and contact details; technical logs and administrator activity logs.
3. Storage locations and recipients
The project documentation specifies infrastructure at Hetzner in Germany (EU), using PostgreSQL and Valkey. This does not mean that all providers listed below process data exclusively in the EU. [CLARIFY BEFORE PUBLICATION] Confirm the actual production configuration, each provider’s regions, agreements, retention periods and grounds for international transfers: [PROVIDER REGIONS AND TRANSFER TERMS].
- Google Gemini: when server-side AI is enabled and consent is given, receives the question, profile and record context allowed by permissions, and any photos or voice attached by the user. Details below.
- Cloudflare R2: is intended for private storage of photos and thumbnails. The client uploads and reads file bytes using temporary signed links; the API stores metadata. [CLARIFY BEFORE PUBLICATION] R2 is not configured on staging; confirm the storage region.
- Expo Push → Apple/Google: when push is enabled, receives the token, event or reminder title (up to 180 characters), time, type, and record and family identifiers. The title may contain personal information you entered; keep lock-screen visibility in mind. [CLARIFY BEFORE PUBLICATION] Delivery to real devices has not yet been confirmed.
- RevenueCat and App Store / Google Play: are intended to process user and purchase identifiers, product, period and subscription status. The API does not accept bank card details. [CLARIFY BEFORE PUBLICATION] Real purchases and store configurations have not yet been verified.
- Telegram: if forwarding is configured, receives the feedback text itself, without added profile or medical context. Do not include unnecessary sensitive data in your message. A Telegram failure does not undo the saved feedback; repeated delivery may create a duplicate. Automatic deletion of text already sent to Telegram has not been implemented.
- Sentry / OpenTelemetry (for example, Grafana): only when enabled, receive sanitised technical details: error type, route pattern, method, status, random request ID, version and duration. The allowed set excludes health text, AI questions, request bodies, tokens, email, SQL and signed links. [CLARIFY BEFORE PUBLICATION] Actual production telemetry has not yet been verified.
The website uses Google Fonts: your browser contacts Google to download fonts. [CLARIFY BEFORE PUBLICATION] The reviewed website and API implementation contains no advertising SDKs, analytics tracking or mechanism for selling data; PostHog has not been implemented. Confirm the operator’s commitment that “we do not sell data” and the absence of advertising-related transfers across all production integrations.
4. AI and consent
We keep a history of your choices concerning personal data processing, acceptance of the medical disclaimer, AI request processing, requests for notification permission and grants of that permission. The value, policy version and time are recorded. A missing record is not treated as consent.
Processing an AI request requires your current consent; without it, the request is rejected before Gemini is contacted. Context respects permissions: names and birth dates of accessible children, pregnancies, permitted medical and dietary notes, and recent health, sleep, recipe and event records. With full access, the last eight successful messages in the current conversation may be added. Conversation history is not added to requests from members with limited permissions. Photos and voice are sent only when attached to the request.
You can change or withdraw consent in the app: [IN-APP PATH: CONSENTS]. Withdrawing AI consent blocks new server-side AI requests, but does not recall a request already sent or automatically delete history. Withdrawing other choices does not itself delete data or block every feature. To stop push notifications, disable permission in the operating system and unregister the device token; changing the consent record alone is not enough. [CLARIFY BEFORE PUBLICATION] The mobile screens for these actions still need verification.
5. Family access
The server checks active membership and section permissions: the owner can write in every section, and the partner can do so by default too. By default, a nanny can read profiles and keep sleep records; medical records, nutrition and AI are unavailable to them. Grandparents can read profiles, sleep, development and organisation by default. Individual permissions may change these defaults, except the owner’s permissions. Inactive members have no access.
Restricting permissions does not erase data already downloaded to someone else’s device. A previously issued file link remains valid until it expires; no new link is issued after access is revoked. [CLARIFY BEFORE PUBLICATION] The family membership management interface still needs verification.
6. Export and deletion
Export provides a JSON file containing your own account, consents, devices without push tokens, and permitted data from active families. It includes metadata of accessible files, not photos or signed links; the owner can access the family’s payment receipts. Internal administrator logs are not included. Path: [IN-APP PATH: EXPORT].
Deleting an account removes the account, sessions, devices, consents and linked personal feedback. If you are an owner, families you own and their data are also deleted; other members’ accounts remain. If you are not an owner, shared family records remain with the author reference cleared. Path: [IN-APP PATH: ACCOUNT DELETION].
Storage objects are deleted asynchronously: the deletion obligation is saved before network attempts; failures are retried at startup and during queue processing. Cleanup is repeated after upload links expire. An account deletion response does not guarantee immediate deletion of every photo; no final deadline has been set in the event of a failure. External requests already sent cannot be recalled.
Minimal payment event identifiers are retained to prevent repeated processing, along with store identity hashes to prevent repeat trials (these are pseudonymous, not anonymous data), storage cleanup receipts, and sanitised audit records containing action, time, role and outcome. For a retained transferred purchase, identifiers follow the new recipient. Deletion of data held by payment providers has not been implemented; deleting an account does not cancel a store subscription.
7. Retention and protection
A single automatic record deletion period has not yet been established. Ordinary deletion of a synced record may retain a technical deletion marker. [RETENTION PERIODS: records, feedback, logs, audit, receipts, hashes and backups] must be approved before publication. Anonymous website feedback is not linked to an account: deleting it requires a separate support request, rather than account deletion.
Data protection measures include sign-in verification, session revocation, separation of families and sections, temporary private file links, request limits and sanitised technical logs. Staff access is role-limited and logged. It covers account and operational details, but not medical records, AI conversations or photos; feedback text is an explicit exception for support.
[CLARIFY BEFORE PUBLICATION] The documentation provides for nightly logical database backups; these do not include R2 file bytes. Production recovery, backup cleanup periods, disk encryption, point-in-time recovery and monitoring have not been confirmed. This draft does not guarantee them. Logging settings across the infrastructure also need a separate review.
8. Requests and changes
For questions about data, contact support or [SUPPORT E-MAIL]. [PROCEDURE AND RESPONSE TIME FOR DATA REQUESTS], the competent supervisory authority and the procedure for notifying users of policy changes must be approved before publication.