Privacy Policy
Last updated: 2026-11-20 · Effective: 2026-11-20
⛔ READ FIRST — SAFETY & SCOPE LIMITS.
Refairly is not a safety system, crisis service, mental-health provider, therapist, counselor, mediator, law firm, arbitration body, medical service, or emergency service. The AI is a software tool, not a professional.
- In any emergency or imminent-danger situation, contact local emergency services immediately. Refairly cannot respond to crises in real time.
- In the couple features, the Service does not perform server-side detection of abuse, self-harm signals, suicidal ideation, threats, or safety-critical content. The AI's safety behavior there is driven only by its system prompts and is not a clinical safeguard.
- In the family features, an automated safety check runs on what a young person writes in her weekly check-in, in her own Refairly chat, in notes on her goals and in her Kitchen chat and table, and on the messages she herself sends to Refairly in a chat (see §10.5); others' messages to Refairly are not checked. It is an aid that shows help resources and, in defined cases, asks a parent to check in. It is not a clinical safeguard and it will miss things.
- Do not use Refairly as a substitute for professional help if you, your partner, or your child are experiencing domestic violence, abuse, coercion, stalking, child-safety risk, self-harm risk, severe mental-health distress, or criminal threats — seek qualified professional help.
- Do not use Refairly to contact a person you are legally restricted from contacting (restraining order, no-contact order, court protection order). Doing so may violate the order.
- You should not use Refairly while in a relationship dynamic involving abuse or coercion. A "shared chat" feature in such a context can be weaponized to surveil, pressure, or control a partner. See §10.10 below.
By using the Service you acknowledge these limits.
⚠ Refairly is not a crisis service.
In the couple features, Refairly does not perform server-side monitoring or content scanning of your messages for safety-critical patterns, abuse signals, self-harm risk, or threats. The AI in those surfaces may surface help resources or stop participating when triggered by its system prompts; this is prompt-level behavior, not a clinical safeguard, and we cannot guarantee detection of any specific situation. The automated safety check described in §10.5 runs on a young person's check-ins, her own Refairly chat, her goal notes and her Kitchen chat and table, and on the messages she herself sends to Refairly in a chat; other people's messages in a chat with her — an adult's message to Refairly included — and chat messages not sent to Refairly are not checked, including when a member's own Refairly reads the chat. It is not a guarantee either.
Refairly is not a 24/7 service and does not provide crisis counseling, crisis management, emergency response, mental health treatment, or domestic violence intervention. No human at Refairly is a licensed therapist, crisis counselor, or emergency responder.
If you or someone else is in immediate danger, call your local emergency number — do not rely on Refairly to detect, respond to, or escalate your situation.
- 🇹🇷 Türkiye: 112 (emergencies, including a risk of suicide), ALO 183 Social Support Line (Ministry of Family and Social Services), and the KADES app (the police app for violence against women)
- 🇨🇦 Canada: 911, 988 (Suicide Crisis Helpline), Kids Help Phone 1-800-668-6868
- 🇬🇧 United Kingdom: 999, Samaritans 116 123, Childline 0800 1111
- 🇺🇸 United States: 911, 988 (Suicide & Crisis Lifeline)
- 🇪🇺 EU general: 112, 116 111 (child helpline, where operated)
These numbers are given for convenience. We check them, but they can change and may not cover your situation; if you are not sure which to call, call your local emergency number.
Refairly is not a substitute for professional mental health, crisis, or domestic violence services. By using Refairly you acknowledge these limitations.
Refairly ("we," "us," or "our") is an AI-powered companion app for romantic couples and their families, operated by Ali Onur Can İrey, a sole proprietor (gerçek kişi) based in Turkey. This Privacy Policy describes how we collect, use, store, and protect your personal data when you use the Refairly mobile application, the refairly.app website, and related services (collectively, the "Service").
This Privacy Policy is a transparency notice, not a contract. Your use of the Service is governed by our separate Terms of Service. By using the Service, you acknowledge that you have read and understood this Privacy Policy. Adult accounts are available only to people aged 18 and older. A young person aged 13 to 17 may use a limited teen account only when a parent or guardian who already uses the Service creates it for them (§10). Turkish-language users should also see our KVKK Aydınlatma Metni.
1. Data Controller
Ali Onur Can İrey (gerçek kişi / sole proprietor) Email: aoci@refairly.app Website: https://refairly.app
For the purposes of the EU General Data Protection Regulation (GDPR), the Turkish Personal Data Protection Law (KVKK), the UK GDPR, Canada's PIPEDA, Quebec's Act respecting the protection of personal information in the private sector (Law 25), and other applicable data-protection laws, Ali Onur Can İrey is the data controller responsible for your personal data. For Quebec users, the same person is the person in charge of the protection of personal information required by Law 25, reachable at the address above.
2. What Refairly Does — Feature & Data Map
Refairly is a shared space for a couple, and for the people its users look after: children, a pet, and a mother or father of one of them. The product is organized into feature areas ("layers"). Each layer collects different data, sends different content to AI providers, and shares different things between you, your partner, and — in the family layer — a young person with a teen account. Knowing which feature does what is the most important privacy concept in this notice. §2.1 is the canonical map; every other section in this policy refers back to it. Across the layers, Refairly — the assistant inside the app — reads what the person who asks it may see and suggests changes as cards; a card changes nothing until that person taps it. In a trip plan's own chat it changes the plan itself, without a card. In an adult's own Refairly chat it also keeps short notes about that adult, which change nothing else in the app and which they can see and delete. §4.7 sets out what it reads and what it can do.
What "private" means here. Where this policy says something is "private", "only you" or "not shown" to someone, it means other people using Refairly cannot see it. It is still processed by our service providers (§5), can be reviewed by us when it is reported or when it is needed for safety or security, and can be disclosed where the law requires (§10.12).
2.1 Feature inventory — data, who sees it, AI providers
Couple layer
| Feature | What you do | What we store | What your partner sees | What is sent to which AI provider |
|---|---|---|---|---|
| Account & profile | Sign up; set display name, language preference, country, city, bio, optional date of birth and gender, optional profile picture | AspNetUsers-managed identity, hashed password (when not social-login), email, country, city, bio, optional date of birth, optional gender, profile-picture key (with auto-generated thumbnail + medium sizes on AWS S3, server-side encrypted) |
Display name + profile picture once you connect | Display name is sent to Anthropic with every couple-context AI call. Refairly, the assistant, reads your account — your email, app language, date of birth, gender, city, country and bio — in your own Refairly chat when it needs it, never in a chat with others (§4.7) |
| Individual onboarding | Short personality / communication-style intake ("Personalize", UserProfile.ProfileData) | UserProfile.ProfileData JSON blob (your free-text + structured answers) |
Private to you — not shown to your partner | Embedded into Anthropic prompts for every Conflict-layer private session + assessment. Refairly, the assistant, reads it with your account in your own Refairly chat when it needs it, never in a chat with others (§4.7). The interests you picked may fill in your taste for a trip plan where your taste profile names none, and so go to Anthropic with that plan's build and chat, whose text everyone on the plan reads |
| What you'd like help with | The first question before sign-up: you and your partner, your children's days, a teen, someone who helps, your mom or dad, a pet (any of them, or none) | User.HelpWith (a set of choices) |
Private to you — not shown to your partner | Sets the app up for you (the setup steps after sign-up and the Start-here card on Today) and is read by Refairly, the assistant, when it answers you; it reaches Anthropic only as part of your Refairly chat |
| Relationship facts | Answer five short questions about the relationship — what you are (married, engaged, dating, living together, or co-parents), how you live, since when, whether you have children, and whether you are expecting a child; one of you answers and the other confirms or changes the answers | Fields on the Relationship row (what you are, how you live, the year you got together, whether you have children, whether you are expecting; who answered and who confirmed, and when) |
Both partners see the facts — they describe you both | What you are shapes the questions the app asks. What you are, how you live, since when and whether you are expecting go to Anthropic with "About us" (§4.1): when it writes the "About us" synthesis, in the couple's Kitchen chat and table, when Refairly answers in the couple's chat, and in a trip plan for the two of you; a family's trip plan carries what you are, how you live and since when, without "About us" or whether you are expecting, and everyone on that plan reads its text. Refairly, the assistant, also reads the facts in your own Refairly chat and can suggest a change by a card you tap, which your partner then confirms (§4.7). The "expecting a child" answer is health data and is stored and used only after your explicit consent (§2.3) |
| Connection requests | Send / accept / decline / cancel invitations — by the other person's display name, or by scanning or opening their invite code or link, which carries their display name | ConnectionRequest rows with sender/receiver ids, optional note, status, timestamps |
Only the recipient sees the request; status visible to both | None for the requests themselves. Refairly, the assistant, reads the requests waiting for you and the one you sent — the names, the dates and any note — in your own Refairly chat, and can send, accept, decline or cancel one by a card you tap (§4.7) |
| Joint onboarding ("About us") | After you connect, each partner answers 5 free-text questions about the relationship; AI synthesizes a shared summary | JointOnboardingResponse (per partner — raw 5 answers as JSON, private to writer); Relationship.SharedContextJson (the AI-synthesized 5-field summary — visible to both) |
The synthesized summary; not the raw answers | Anthropic (one synthesis call). The synthesized "About us" is also embedded into nearly every later AI call as couple context |
| Daily Reflection | One rotation-question per day; you write a free-text answer | DailyCheckIn row (UserId, RelationshipId, LocalDate, QuestionSlot, Answer, SubmittedAt) — private to you |
Nothing. Your reflections are not shown to your partner | None unless you opt in to Reflection Insights (§4.6). With that consent, your recent reflections are distilled by Anthropic into a private digest that is used inside your own private AI conversations; a rough score drawn from it also feeds the trend label described under the relationship check below, which carries no words from your reflections. Refairly, the assistant, does not read the reflections or the digest (§4.7) |
| Family pulse | Set a one-word mood per day, and choose who sees it | PartnerPulse row (UserId, RelationshipId, LocalDate, Pulse token) + PulseAudience (who may read it) |
From the start your partner sees it — a co-parent too — and a parent and each teen they parent who has her own account see each other's; where you choose who sees yours you can stop it for any of them, and a teen can also show hers to a brother or sister who has an account. When only one other person shares your household — the two of you, or a parent on their own and one teen — your word always goes to them. Everyone who sees it is notified when it changes ("Ada: tired"), and your partner's home-screen widget, if they added one, shows it. A caregiver is not among them. Explicit crossing of the privacy wall | None for the pulse itself. Refairly, the assistant, reads today's words you may see and can set your own word, or who sees it, by a card you tap (§4.7) |
| The couple's chat | The chat between the two partners, part of the one chat system (the family's chats are in the family layer below) | DailyChat (kind couple), DailyChatMessage rows (author, body, an optional quoted reply, the AI-turn state) |
Both partners see every message, including Refairly's replies and cards | Messages are not sent to the AI when they are written. They reach it when either of you asks Refairly: with @refairly here, the recent conversation and what both of you may see in the app go to Anthropic with the request (§4.7), and Refairly answers once in the chat and may add cards that only the one who asked can tap; in your own Refairly chat, it can read this chat when it needs to (§4.7). (An AI that reads every message of this chat exists in the code and is switched off; this row and §4.1 change before it is switched on) |
| Growth Plan — questionnaire | Both partners answer 5 free-text questions — about the two of them or, once both are parents of a child on the record, about the household. With children this one plan is the household's plan; each child also has their own list of goals (below). AI writes the 7-section reading and the household's first goals in one call. A parent on their own answers five questions of their own, and the plan is woven from those alone. | GrowthPlanResponse (per partner and per question set — raw answers as JSON, private to writer); the plan's 7-section reading (shared); the first goals (see Goals) |
The reading and the goals; not the raw answers | Anthropic — one synthesis call producing both the reading and the first goals. For a household, each child's name, age and the note the parents keep on the record go with it while every parent the plan is for has given the children's-data consent (§2.3) — not observations, the health list or anything a young person wrote |
| Goals | Goal items, each with one owner: the household (the couple, or a solo parent), one adult (their own items), or one child. They are shown together in one table, which can be filtered by person. A goal item has a title, a description, a category, one-time or ongoing, when (or a date) and its progress — a status, or for an ongoing goal one of five words. Each has a page of dated notes, photos and files, each showing who added it and when, and can carry alarms (§2.10). The people who keep an item — the household's members on the household's, the adult and, in a couple, their partner on an adult's own, the child's parents on a child's — can mark it Parents only; an item a young person wrote on her own list stays hers and cannot be. The household plan's first goals, and the household goals written before this switch existed, are Parents only until someone who keeps them changes that. Goals replaced the plan's steps, the steps about a child and skills; the plan's own chat was removed with them, and the conversations it held were deleted | Goal (title, description, category, kind, when, target date, progress or stage, whether it is Parents only, its place on the household list, who wrote it, who changed it last, whether Refairly wrote it), GoalNote (text up to 1,000 characters and/or one file — JPEG, PNG, WebP or PDF, up to 15 MB — with its name, type and size, who added it and when). Files are kept on AWS S3, private, and served only through short-lived signed links |
The household's items: its members — both partners, or the solo parent — can see, add, change and delete them. An adult's own items: every adult has a list of their own from the start (a caregiver's account has none) and can see, add, change and delete its items; in a couple, the other partner can too. When the couple ends each keeps their own, and the former partner no longer sees them. Two people connected as co-parents share the household's items but not each other's own. A child's items: the child's parents and the child herself once she has her own account; each of them can see, add, change and delete goals, notes, photos and files, and she does not see an item marked Parents only. A young person with her own account also reads the items of the household her parent belongs to, and its adults' own items, that are not Parents only, and changes none of them; she does not see her brothers' and sisters' items. A caregiver given the child reads that child's items that are not Parents only, with their notes, photos and files, and changes nothing on them; she does not see the table. An item's notes, photos and files go to whoever sees the item, and an item marked Parents only is absent for everyone else, not shown as locked. A paused account reads and cannot change anything. Goal photos and files do not appear among a child's papers | The household's first goals are written by Anthropic with the plan's reading (§4.1). Refairly, the assistant, reads goal items and their text notes for the person who asks — in a chat, only those every member may see — reads a goal's photo or file when it needs one to answer, and can add, change or remove goal items and text notes by a card the asker taps (§4.7); an item it adds to the household's list or an adult's own is Parents only unless the card says otherwise. A young person's own Refairly reads her own list, not the household's items or her parents' own. Files on a child's goals need the parent's children's-data consent (§2.3). A young person's notes go through the automated safety check (§10.5) |
| Trip Planner | Anyone in a household starts a plan — a dinner, a day out, a weekend or a short trip — and says who is coming; a young person suggests one, with a parent on it, and a parent says yes. The plan is built at once from the trip's answers and the taste profiles of the people coming. For each person the planner may write one line of things to plan around (food rules, allergies), for that trip only; for another adult, only what that adult asked to pass on; for a child, a parent who has given the children's-data consent sees the line filled in from the child's card (what the child eats, and allergies and intolerances on the health list) and can change or clear it. Each stop comes with up to three real places to choose from; anyone on the trip with an account can pick another, and a parent can change the stay. In the plan's chat Refairly reads and answers every message when no young person with her own account is on the plan; with one on it, it answers only when someone writes @refairly, or while its own question before a build is open, and the other messages stay there as wishes for the others. Refairly changes the plan from the messages it answers, as far as the message's writer may and without a card; on a plan with no young person with her own account on it, that can follow any message, including saving a person's line of things to plan around (§4.7) | TripPlan (city, country, dates, budget, the stay, the planner's wish, and for a couple what the trip is for); TripPlanParticipant (who is coming, a child's rough age band if no birth date, the line of things to plan around); TripPlanContent (versioned plan: the stay and its choices, days, stops and their choices, who picked what, routes, photos); TripPlanChatMessage (on a person's message, also which of Refairly's plan tools ran for it, whether each worked and, when one did not, the planner's short reason — never the words of a line of things to plan around); TripPlanItemInsight (Refairly's notes on a place, one per place in a plan, written when someone asks); TripPlacePhoto (one stored photo per place with its author's credit, shared by every plan that shows the place and fetched again from Google after 30 days; no personal data) |
Everyone on the plan with an account, and the parents of a child on it, see the plan, its choices and picks, the wish and every chat message; the record of what Refairly's tools did is shown to nobody in the app, and the data export gives it to the message's writer (§9.1). A line of things to plan around is seen only by the adult it is about, and by the parents of a child it is about; it goes to Refairly's Trip Planner (its AI) for that trip only and is deleted when the plan ends. A caregiver on the plan sees none of it | See §4.1 (Trip Planner uses Anthropic + non-AI external APIs) |
| Taste profile | Each person answers a few questions about what they enjoy; a parent answers for a child with no account. Used to plan trips for the people on them | ExperienceTasteProfile |
Only you (a child's: her parents). Trip plans are made from it, and the plan's text may say what someone enjoys | Anthropic, as part of a trip plan's build and chat. Refairly, the assistant, reads your taste answers, and a child's you answer for, in your own Refairly chat, and can change them by a card you tap (§4.7) |
| Relationship check (satisfaction) | Now and then each partner answers four short questions about how satisfied they are with the relationship | CoupleSatisfactionSample (your four answers and their total, per round), MeasurementWindow (when a round is due). Also used, without names, to measure on our internal dashboard whether the Service helps couples: how many answered, averages, and how scores move over time |
Your answers and scores are not shown to your partner | Your answers are not sent to the AI, and Refairly, the assistant, does not read them. A coarse trend label (improving or strained), worked out without AI from both partners' totals, how your Kitchen check-ins went and a score from each partner's Reflection Insights digest (where that partner opted in, §4.6), is given to the AI in the Kitchen chat and in the couple's chat to set how readily it speaks; the label carries no answers and no words |
| Conflict (major mediation, "Kitchen table") | For a couple, or for a parent and a young person (13–17) — she consents on her own sheet (§2.3), her words are read by the safety check before they are stored (§10.5), the parent's pool pays, and the parent's couple profile, "About us" and ledger do not enter her table. Five-phase guided flow: each partner describes the conflict → each holds a private AI conversation → both confirm readiness → AI writes one proposal, and a new version after "Not quite" → both sign + follow-up. Requires your one-time explicit consent before first use (§2.3) | ConflictSession, ConflictDescription (per partner), PrivateSessionMessage (per partner — private), ComplaintAssessment (per partner — AI-derived behavioral assessment, not shown to either user, kept for the life of the relationship), ResolutionProposal (shared: each version's text and action items, with each person's answer; the private note to each partner is theirs alone), ResolutionAgreement (shared, with each partner's signature timestamp), ConflictTask (per partner — assigned follow-up), ConflictSharedSummary (per partner — initially private, becomes visible to partner after that partner approves it), ConflictPrivateSuggestion (per partner — private to that partner, not shared), ConflictCheckIn (post-resolution outcome) |
Each partner's description text and private-session messages are not directly displayed to the other partner via any API. AI-generated joint analyses, the proposals, agreement text, and signatures are shared. Once a partner approves their own shared summary, that summary becomes visible to the other partner. Important caveats: both partners' raw descriptions ARE sent to Anthropic in every later AI call for the lifetime of the session; the AI-written proposals necessarily reflect substance from both partners' descriptions | Anthropic for every Kitchen AI call. Payload includes both partners' display names, raw descriptions, full UserProfile.ProfileData, the relationship's "About us", the calling user's complete private-session transcript, your Reflection Insights digest if you opted in, and (for the proposal, the check on a "Not quite" request and the closing package) both partners' AI-generated ComplaintAssessment JSON. A parenting Kitchen table may be tagged with a child; the tag decides which parents may take part, and the child's name and age go to the AI with the household context — nothing else from the child's record |
| Quick Repair (lightweight conflict, "Kitchen chat") | Either partner — or a parent and a young person, either of them — starts a short shared AI-mediated thread; AI is brain-gated and only speaks when there's value. Requires your one-time explicit consent before first use (§2.3) | QuickRepair thread; QuickRepairMessage (per message — both user and AI; AI replies attributed to "Refairly", not a user) |
The two people see every message including the AI's. A young person's message that trips the safety check is kept for her alone and pauses the chat (§10.5). Her other parent can see that a chat or a table exists, with whom, its status and a signed agreement — not what anyone wrote in it | Anthropic, via the Refairly Brain. The whole transcript + display names + the couple's "About us" — or, for a parent and a young person, the family block (who is the parent, her name and age, the parents' situation) — sent each turn. A parenting Kitchen may be tagged with a child; the tag decides which parents may take part, and the child's record is not sent to the AI |
| Buzz (haptic ping) | Send a short haptic vibration envelope into one of your chats — the couple's, a parent's with a young person, or the family group | Buzz row (chat, sender, envelope JSON, idempotency key, status, played-at) |
Everyone else in the chat sees and plays it; in a chat of two you see it was played | None |
| Buzz Calls (live haptic call, with voice) | Phone-call-style live session between the two people of a two-person chat — the couple, or a parent and a young person. The vibration goes through a real-time hub; when the microphone is allowed, the two of you also hear each other (§2.11). A call to a young person rings through quiet hours | BuzzCall row (chat, caller, callee, whether it was a voice call, state, started/answered/ended-at, end reason, duration, idempotency key). Intensity samples and call audio are not stored. |
The other person sees the call attempt, the state changes, and the duration in history | None. Intensity samples relay through our SignalR hub only — not stored, not logged, not pushed to any provider. Call audio goes between the two phones and is not sent to any AI (§2.11) |
| Subscriptions & token packs | Every new household starts with a welcome week: for seven days everything is open, the AI included, on one allowance the household shares, with no plan to choose and no payment step at sign-up (§12). Buy a subscription (Weekly / Pro / Max) — in a couple, one partner's covers both — or a consumable token pack through Apple App Store or Google Play | UserSubscription (tier, store transaction id, original transaction id, store, product id, store expiry, period start/end, computed token limit, computed bonus tokens, computed usage, and when the welcome week ends). TokenUsageLog (per AI call — UserId, token count, estimated USD cost, provider, model, request type — no prompt content). WebhookEvent (audit of store webhooks). SubscriptionAuditLog (state transitions) |
Partner sees that a subscription is active for the couple + which partner is the payer (payerUserId); with two subscriptions, each one's plan and renewal date, and a paying partner sees each person's share of use as a percentage; not your payment instrument |
None (these systems don't talk to AI) |
| Notifications | Push + in-app inbox (email is transactional-only) | DeviceToken (per device — FCM/APNs token, VoIP token if iOS, platform, device id/name, OS version, app version, last used). NotificationPreferences (master push and email switches, quiet hours kept as local times, the time zone your phone reports when you sign in and when you come back to the app, per-type push toggles). NotificationHistoryEntry (in-app inbox — Type, Title, Body, full Data dict, IsRead/IsSent flags, per-channel delivery flags). Your phone's time zone sets your quiet hours, the evening reflection reminder, the day your word of the day is filed under, the time each alarm rings for you, the times on your own day (§2.12), and what Refairly takes as today and now for you (the zone's name goes to Anthropic with each request you make of Refairly). It is also where the day of someone you add — a child, Mom or Dad, a pet — and the Home day you first open start their clock, which can then be changed on that day; a day that had no clock of its own was given, once, the time zone a phone of one of its keepers (for the Home day, one of its members) had reported |
Notifications fired by partner actions reach you per the feature flows above; preferences are private. Push payloads may carry short content previews (e.g., a chat message snippet or your pulse value) through the delivery infrastructure — except for any chat that includes a young person, where pushes carry no message text (§10.5). A notice that a day changed never names a medicine | Push delivery: Firebase Cloud Messaging (for both Android and iOS regular pushes) + Apple APNs (for iOS VoIP pushes only) |
| Safety reports | Report a message or a person from any chat surface, a reply from Refairly, or a note on a goal | SafetyReport (who reported, which message/surface, the reasons chosen, your optional comment, your IP address and device user-agent at the time of the report) |
Nothing. The reported person is not told who reported them | None. Reports are read by a person (§10.11). A copy of each report — with your account id, the message or surface reported, the reasons, your comment and your IP address — is emailed through Resend (§5.4) to our support mailbox and to a second mailbox of ours at Microsoft's Outlook.com, so a person can review it |
| Report a problem | Tell us something is not working, from Settings | What you write (up to 2,000 characters), your account role and, if you leave the switch on, basic device details (app version, operating system, model, language) | Nothing | None. It is emailed to our support mailbox through Resend (§5.4) |
Family layer (available to couples who choose it, and to someone on their own who looks after a child, a pet or a mother or father; every adult has their own day in it from the start; §10 explains the rules for children and young people, §10.15 for a mother or father)
| Feature | What you do | What we store | Who sees it | What is sent to which AI provider |
|---|---|---|---|---|
| Children as records | A parent adds a child: name, date of birth, optional gender, and the child's card — one note in the parents' words, what the child eats, an emergency contact and a health list ("what to watch for": allergies, intolerances, conditions and other things, each of which can be marked serious). The parents also write observations in care about the child's wellbeing: notes on the child's "Notes and papers" list, kept to the parents, added and not changed afterwards, and — like a paper — written only while the parent's children's-data consent stands. Requires the parent's one-time explicit consent before the first child is added (§2.3) | Person (display name, birth date, gender, note, what the child eats, emergency contact, the child's time zone, account state), HealthEntry (each item, its kind, whether it is serious), Keeper (the child's two parents), KeeperNote (each observation: its text, who wrote it, when) |
Both parents of the child see the whole record. A partner who connects later becomes a parent of every existing child and sees the whole record too; a teen with an account is told. A caregiver the parents give the child to sees the card, not the observations (§10.13). A child without an account sees nothing; a young person with an account sees her goals, her Day, her pocket money, the papers meant for her and the course logs of her medicines that are not Parents only, and not her card, its health list or a parent's observations about her | The child's name, age and the note go to Anthropic when the household's Growth Plan is woven, while every parent it is woven for has given the children's-data consent (§2.3, §4.1). Refairly, the assistant, reads the card — the health list included, which is health data — and the observations for a parent of the child who asks, in their own Refairly chat or in a chat whose every member is a parent of the child, and can suggest changes to the card by a card the parent taps (§4.7). What the child eats, and the allergies and intolerances on the health list, are offered to a parent planning a trip with the child, who can change or clear them before they go to the Trip Planner for that trip |
| Your own day | Every adult has a day of their own from the start, in a couple or not (a caregiver's account has none), and keeps it when a couple ends or the two of you become co-parents. On it: day items you add, or in a couple (not co-parents) your partner adds — a meal, a walk, an appointment, and medicines, which start when you add your first one yourself: that is your own-health consent (§2.3, §2.12) — each linked to one of your goal items if you choose, and who ticked an item done or recorded a medicine as taken or not taken. Your own day follows your phone's time zone (§2.12) | DayItem and DayOutcome rows on your own Person row (kind adult), with who made each change; reminder state |
You and, in a couple (not co-parents), your partner, all of it (otherwise only you): there is no item one of you keeps from the other. Your partner may add, change and tick your items; a medicine deleted from your day, by either of you, is deleted for good (§2.12). Your children, a young person with her own account and anyone who helps look after someone in your family do not see it. When the couple ends, or the two of you become co-parents, each of you keeps your own day and the other's access ends at once | Refairly reads your day — your medicines only while your own-health consent stands — for you or your partner, in your own Refairly chats and in the couple's chat, never in a chat with a young person or a caregiver in it, and can change it by a card the asker taps (§4.7). It does not advise whether a dose may be taken |
| Mom or Dad, looked after by the family | A son or daughter adds their mother or father: name, date of birth, whose parent they are, and their card — a note, what they eat, who to call; their day; their goal items; notes; and, once they have agreed, their health list, medicines and papers. Before adding them, the son or daughter confirms a statement (§10.15). Their health list, medicines and papers stay locked until their son or daughter records that they agreed in person, and Refairly reads nothing of their record until they have agreed to that too — each answer with who recorded it, when, and the words they were shown (§10.15) | Person (kind elder; no account), Keeper (their own son or daughter only), HealthEntry, DayItem and DayOutcome, goal items, notes and papers, Caregiver links, and the record of their answers (on their Person row and in the consent audit log) |
Their son or daughter, and that person's partner through the partner link while the couple lasts (not co-parents); the partner may tick their day and is reminded of their medicines only after choosing "Ring me too". Notes and items marked Parents only are the two partners'. Someone the family asks to help look after them sees their card, day and goal items — and their health list, medicines and papers once they have agreed — and not notes, anything marked Parents only, the record of their answers or who else helps. A young person at home sees their goal items in the household's table unless they are marked Parents only, and nothing else about them. When the couple ends they stay with their own son or daughter, even when that is the one who leaves, and the other partner's access ends at once | Refairly reads nothing of their record until they have agreed in person to its assistant reading it; what someone writes about them in a chat is sent as that person's words. After that, Refairly reads what the person who asks may see of them — their card and health list, their day and medicines, their goal items, the notes, a paper when it needs one — for their son or daughter or that person's partner, in a chat only where every member is one of those two, and suggests changes by cards the asker taps (§4.7); what it needs is sent to Anthropic for that answer |
| A pet | A keeper adds a pet — someone on their own, or either partner: a name, what animal, its card (a note, what it eats, who to call — the vet), its day (walks, meals, medicines, appointments), notes and papers (a vaccination card). No consent is asked: a pet's record is kept as part of its keepers' account. In a couple (not co-parents) both partners keep the pet, and a partner who connects later becomes a keeper too. Someone who helps can be given just the pet | Person (kind pet), Keeper, DayItem and DayOutcome, notes and papers, Caregiver links; an expense can be for the pet |
Its keepers; a caregiver given the pet; and a young person at home whose parent keeps it, who sees and ticks the pet's day except items marked Parents only; her tick shows her name, as every tick shows who made it, and the app tells her so. When a couple ends, the pet goes with whoever keeps the family; a couple with only a pet chooses who takes it | Refairly reads and changes the pet's card and day for its keepers, by cards they tap (§4.7) — in a chat only where every member keeps the pet, so never in a chat with a young person or in her own Refairly chat; a vet's details typed under "who to call" go with it when needed |
| Expenses & budgets | Record expenses for the two of you, the house or one or more of the people you look after — children, Mom or Dad, a pet: amount, what it was, a category (suggested on our servers from the words you wrote, using a fixed word list — no AI), who paid (either of you, or shared money, which does not move the balance), split, an optional receipt photo, recurring entries, soft budgets; settle up a month at a time ("who owes whom", then the month is cleared); export a PDF ledger. A top-up a parent gives a child as pocket money is also written here as a household expense ("Pocket money for {name}": the child's expense, paid by that parent, split half and half; the split can be changed in Money); the two are linked, so changing or removing either one changes or removes both | Expense (with who added it and, for a top-up, the pocket-money entry it mirrors), ExpenseBudget, receipt images on AWS S3 (private, served only through short-lived signed links) |
Both partners, or the solo parent. A line names the person it is for only to someone who can see that person now, or who kept them once; to anyone else it says "someone" | Refairly, the assistant, reads this month's expenses and budgets in a partner's or a solo parent's own Refairly chat, and in the couple's chat, and can add, change or remove an expense or set a budget — and, for a couple, settle a month — by a card the asker taps (§4.7). It does not read receipts. "Talk it through in the Kitchen", shown only on an expense for a child, only opens a Kitchen conversation with a pre-filled description that you can edit before anything is sent |
| Family map | See the household on one screen: members, life stages, and what needs attention | Nothing new — it is assembled from the records above | Both partners | None |
| Teen account — claim | A parent's phone shows a one-time code, or the parent sends her a one-time link that works for 24 hours; the young person redeems it on their own device and creates their account: email address (may be a parent's), password, display name; they read the teen contract and decide whether their check-ins may be turned into themes | ClaimCode (hashed; a code expires after 15 minutes, a link after 24 hours), the teen's User row (role Teen, linked to the child record), the teen's consent record (§2.3) |
Parents are notified that the account was claimed. The teen's email and password are theirs; parents do not see them | None |
| Teen — weekly check-in | The teen answers a short weekly check-in across three areas of life — school, home and outside ("street") — a mood and optional text per area; every area can be skipped. She can switch on "share my plain words" in her settings | ChildWellbeing rows, her entries (with whether her parents see its mood number), ChildTheme (one short theme per area per week, with a conversation opener and a "don't" for parents), the moment she turned sharing on |
Her parents see the mood numbers she picks (her screen says so when she answers). They see her words only while "share my plain words" is on — what she writes from the moment she turns it on; turning it off hides every answer from them again, and what she wrote before she turned it on stays hidden. Answers the safety check flagged are shared like the rest while it is on, and her screen says so before she turns it on. With the teen's consent parents also see the weekly themes ("wants calmer evenings"), not a quote. Home themes are phrased as the young person's needs, not as accusations. The teen sees every theme that was shared, on their transparency log | With the teen's consent, the week's entries are distilled by Anthropic into themes. Every entry also goes through the automated safety check (§10.5), consent or not. Her own Refairly chat reads her recent entries, her words included; a parent's Refairly does not read her check-in words, shared or not — it reads the themes she agreed to share and her mood numbers (§4.7) |
| Teen — transparency log ("what my parents see") | The teen reads exactly what was shared about them | Derived from ChildTheme and TeenCrisisEvent counts |
The teen. It lists every theme shared, how many entries they wrote, and how many times the safety check showed them help resources | None |
| Teen — pocket money | Parents and the teen record pocket-money entries: money given, money spent and money she received from someone else. A top-up a parent gives is also a household expense (see Expenses); what she spends and money she records as received herself are not | PocketMoneyEntry (amount, currency, note, who created it, when) |
Parents and the teen; a caregiver given the child reads the entries, changes none of them, and does not see the household money a top-up appears in | Refairly reads her entries and balance for her parents and for her, and can add, change or remove an entry by a card the asker taps (§4.7) |
| The Day | Parents keep each child's day (the same Day holds your own day, Mom or Dad's and a pet's, in the rows above): repeating and one-date items — school, a course, a meal, sleep, a walk, an appointment, a pick-up, medicine (the dose as the parents wrote it and its time, or for a medicine given as needed the gap between doses and the daily limit) and other things — and who ticked an item done, recorded a medicine as given or not given, with an optional note. Medicine items remind the people who look after the child that day, and any item can carry alarms (§2.10). Parents can mark an item Parents only. A young person with her own account adds her own items (not a pick-up, a medicine given as needed or a monthly item), changes, skips, pauses, stops and puts back her own, ticks the items she sees, and records a medicine as taken or not taken. Each medicine has a course log of the doses recorded on it; anyone who reads the log can make a copy of it for a doctor, without who gave each dose; the phone builds it. The Home day is the household's own day, for chores and bills: its members add items (never a medicine), can say who does each one — a member, a young person at home, or anyone — and can let an item stay on the day until someone ticks it, shown with the date it has waited since | DayItem (with its changes from a date, one-date changes, skipped dates and pauses; on the Home day, who does it and whether it stays until ticked), DayOutcome (what was recorded, by whom, when, any note), reminder state; the Home day itself is a Person row keyed to the household, holding its clock |
The child's parents — all of it, including the items a young person adds herself (medicine too) and every tick she makes. A young person sees the items her parents set, except those marked Parents only, and her own, with every record on them and the course logs of the medicines among them. A caregiver given the child sees the child's day for today as the young person sees it — every item except those marked Parents only, the young person's own items (medicine too) included — with every record on them, and the course logs of the medicines she sees; while a parent leaves that switched on she also adds items for today (not a medicine), changes or skips today's items and ticks them, and the young person's own items stay hers. The Home day: its members see and change all of it; a young person with her own account whose parent is a member sees its items that are not Parents only, on any date, and ticks those given to her or to anyone, changing nothing. An item marked Parents only can be given only to a member. A caregiver does not see the Home day, and it holds no health data | Refairly, the assistant, reads the Day as the person who asks sees it — in a chat, only what everyone in it may see, so in a chat with a young person never an item marked Parents only — and can add, change, skip, pause or stop items, medicine included, and record a dose the person says was given, by a card that person taps (§4.7). It reads and changes the Home day only in a member's own Refairly chat and in the couple's chat, never in a chat with a young person or in her own. It does not advise whether a dose may be given. Caregivers have no Refairly |
| Alarms | An alarm on a Day item or a goal rings like the clock app on the phones of the people it is for — once on a date, on chosen weekdays, a number of days before a goal's date, or at a Day item's own time — with an optional heads-up before it and an optional label. Whoever may change the item or goal can set one for anyone who can see it; anyone who can see it can set one for themselves alone. An alarm for someone else is a request: it rings them only after they accept it. Each phone schedules its own alarms (§2.10) | Alarm (what it is on, when it rings, the heads-up, the label, on or off, who set it), AlarmPerson (each person it is for: who asked them, their answer — waiting, accepted or declined — and when, whether it is on for them, their own heads-up). Our servers do not record whether an alarm rang, was snoozed or was stopped |
Everyone who can see the item or goal sees its alarms: when each rings, who set it, whom it rings, how each of them answered and their own settings. An alarm rings only people who can see what it is on | Refairly, the assistant, reads the alarms on an item or a goal for the person who asks — in a chat, only on what every member may see — and, in that person's own Refairly chat, the requests waiting for them and their next alarms; it can set, change or remove an alarm, change your own settings on one, or accept or decline a request, by a card the asker taps (§4.7). Caregivers have no Refairly |
| Family chats | One chat system for the household: a family group for the parents and every young person with an account, a chat of each young person with all her parents, and a chat of each parent with her, just the two of them (where their buzzes and calls live). Refairly answers only when someone writes @refairly, once per request, and helps a young person who asks it to "help me say it" | DailyChat (kind, members, whether a minor is present), DailyChatMessage (author, body, AI-turn state) |
Every member of a chat sees every message, including Refairly's replies and cards. A parent outside a two-person chat does not see it; a young person sees nothing from before the day she joined a chat | Only when someone writes @refairly (or a young person asks for help saying something) does the recent conversation go to Anthropic, which replies once, may add cards that only the one who asked can tap, and runs the safety check on it when a young person sent it (§10.5). In a chat, Refairly reads and shows only what every member may see (§4.7). Other messages are not sent to the AI when they are written and are not checked; a member's own Refairly can read a chat they are in when they ask it, and a chat with a young person only while every young person in it has her teen consent active (§4.7). If a young person in the chat has not consented to AI replies, the AI does not reply in that chat; her own tagged message is still checked, and an adult's is not answered. The AI turn is paid from a parent's pool, not the teen's (§12) |
| Themes into the parent's chat with her | After a check-in, the theme for that week is posted into each parent's chat with her as a conversation starter | A DailyChatMessage posted as a system message |
The parent and the teen in that chat | Produced by the check-in distillation above; no extra AI call |
| Weekly family digest | Parents receive a weekly summary: a few themes, one conversation opener and one "don't" for each | Assembled from ChildTheme; posted as a system message into the family group chat and sent as a notification |
The child's parents, equally | None — assembled from the week's themes without an additional AI call |
| Safety check in family surfaces | Runs automatically on a young person's check-in entries, on what she writes to her own Refairly chat, on her notes on goals, on the messages she sends to Refairly in a chat (with @refairly, or in a trip plan's chat while Refairly's question before a build is open), and — before it is stored — on what she writes in her Kitchen chat or at her table, where a signal pauses the chat or the table and keeps her words for her alone | TeenCrisisEvent (which area of life, the AI's short description of the signal, whether resources were shown, whether parents were asked to check in). The description is blanked after 90 days and the record deleted after 365 days |
The young person is shown help resources. Parents receive a contentless "please check in with them" for signals about school or outside life, not for signals about home. The description is seen only by us, on a report or a legal request | Part of the AI calls in the rows above (Anthropic) |
| Your Refairly chat (an adult's) | Your own conversation with Refairly, the assistant. It answers every message. It reads what you may see in the app, as you ask (§4.7). When you ask for a change, it shows the change as a card; nothing changes until you tap the card, and the tap does what the app's own screen does, with your rights. It also keeps short notes about you from this chat, by itself and without a card — lasting things you tell it about yourself, even in passing (how you like to be called, your work and its hours, your routine, what you like or avoid), and what you ask it to remember — in app versions that show those notes, and reads them in its later answers to you; you see and delete them in Settings → What Refairly remembers (§4.7) | PrivateChatMessage (your messages, Refairly's replies, the AI-turn state; on each of your messages, the app's record of that turn — which reads Refairly ran and whether each worked, never what they returned; on a reply, whether its turn kept a note), RefairlyAction (each card: what it would change, its state — suggested, done or dismissed — and when), RefairlyNote (each note Refairly keeps about you: its text, and when it was kept or last changed) |
Only you — not your partner, not your children. The notes are shown only to you, in Settings; the record of each turn is shown to nobody | Anthropic: your message, the recent conversation with the app's record of Refairly's earlier turns, Refairly's notes about you, the trip plans you are on that were made in the last seven days, and what you may see in the app, read as needed (§4.7). It does not read your partner's private answers or Refairly chat, anyone's private side of a mediation, a young person's check-in words (even those she shares with you), your daily reflections or their Reflection Insights digest, a Kitchen private session, satisfaction-check scores, or the caregiver chats |
| A young person's Refairly chat | Her own conversation with Refairly, on the same surface. It reads her own data (§4.7) and can suggest changes — her goals and notes, her Day and the alarms on what she sees, her pocket money, her pulse word and who sees it, her taste answers, a message sent as her, notification settings, her name and language, turning off "share my plain words" or the themes for her parents, and clearing this chat — by cards she taps, while her teen consent is active | PrivateChatMessage (her messages, Refairly's replies, the AI-turn state; on each of her messages, the app's record of that turn — which reads Refairly ran and whether each worked, and whether the help resources took the place of a reply (removed after 365 days), never what a read returned), RefairlyAction. Refairly keeps no notes about a young person |
Only her — not her parents. The safety check (§10.5) runs on what she writes; a parent it alerts is not told what she wrote | Anthropic — see §4.7 and §10. It reads her recent check-in entries, her words included, and the themes she sent. It does not read the household's plan, goals, Home day, money or agreements, her parents' own goal items, her parents' observations, her card, her brothers' and sisters' records, papers kept to her parents, or her chat with her caregiver |
| A child's papers | Parents add files about a child — a health report, a letter from school, a photo — and choose per file who may open it: the family — with the caregivers they gave this child — or the parents alone (Parents only). Requires the parent's children's-data consent (§2.3), whose text names health reports | Paper (the file's name, type, size, an optional note from the parent, who may open it, who added it) and the file itself on AWS S3, private, served only through links that expire in ten minutes |
The people the parent chose for that file. A young person with her own account and a caregiver given the child read the ones meant for the family; a file kept to the parents is absent from everyone else's list, not shown as locked | Refairly, the assistant, lists the papers the person who asks may open and reads one when it needs it to answer — an image (JPEG, PNG or WebP) or a PDF; it does not read Word or HEIC files. It reads only a file the asker may open, and in a chat only one every member may open; a young person's Refairly does not see papers kept to her parents, or their names. A parent can change who a paper is for, or remove it, by a card (§4.7) |
| The child's card, for a caregiver | The card the parents keep on the record (above) — the note, what the child eats, the emergency contact and the health list — is what a caregiver they give the child to is shown | Fields on Person; HealthEntry |
Both parents, and a caregiver they have given this child. Not the child, not a young person with her own account, not another family's caregiver. The health list is health data — kept for this record and shown to the people above | Through Refairly, the assistant, for a parent who asks (§4.7), and the note with the household plan's synthesis (§4.1). Caregivers have no Refairly |
| Caregiver account — join | A keeper's phone shows a one-time six-digit code, in person, or the keeper sends a one-time link that works for 24 hours, for the first person she will look after — a child, a pet, or a mother or father; the adult redeems it on their own device and creates their account: email address, password, display name, and their own date of birth, which must make them 18 or older. A caregiver the family already has, or had, needs no new code: she can be given another person with one tap, from a list on that person's page (§10.13). A caregiver who already has an account adds another family with a parent's code or link the same way, from inside her app (§10.13) | ClaimCode (hashed; a code expires after 15 minutes, a link after 24 hours; marked as a caregiver code), the caregiver's User row (role Caregiver), Caregiver (which person, who gave them, when, whether she may change the person's Day; once a person is taken back, the row is kept, marked as removed, as the record that she looked after them before) |
The person's keepers are notified once the caregiver confirms their email. When a keeper gives someone to a caregiver the family already has, she is not asked first: she is told, and so are the other keepers. The caregiver's email and password are theirs; the family does not see them. A caregiver is connected to a child only once her email is confirmed | None |
| Caregiver chat | The keepers and one caregiver, about the people she was given — practical notes, both ways. A chat about Mom or Dad is between their son or daughter and the caregiver, not the couple's, so a couple ending neither moves nor closes it | DailyChat of kind caregiver (not marked as containing a minor), DailyChatMessage |
The keepers and that caregiver. No child or young person is a member. It opens with the first person she is given and closes with the last one taken back | None. Refairly does not take a turn in this chat even if someone writes @refairly, the app does not offer it, and it does not read the chat for anyone |
| A young person's chat with her caregiver | When a parent opens it from the caregiver's screen, a young person with her own account and a caregiver she was given have a chat of the two of them; a parent can close it at any time | DailyChat of kind teen and caregiver, DailyChatMessage |
The two of them. Her parents are not members and do not read it; they decide whether it exists | None. Refairly does not take a turn in it and does not read it for anyone |
| Caregiver — her app | Her home has one line for each person she was given — a child, a pet, or a mother or father — with that person's day as she reads it, marked "To check off" where a tick of hers waits, her chat with the keepers, and a bell with her notifications. A line opens the person's page: the card above, the person's Day, goals, a child's pocket money, and the papers and course logs she may read. In Settings she reads her scope written out and the families she looks after, adds a family with a parent's code, leaves one, downloads her data (§9.1) and reads the legal documents | Nothing new — assembled from the rows above; her notifications are kept in her in-app inbox (NotificationHistoryEntry, §8) |
Her. The app names what she cannot see and names exactly the people she was given | None |
| Teen — status line for parents | Parents see that a check-in happened this week, and one notification a week saying so | Derived from the existence of ChildWellbeing teen entries in the last seven days; no new row |
The child's parents. The fact of the check-in; what else they see of it is set out in the check-in row above | None |
2.2 Account information
| Data | Required | Purpose |
|---|---|---|
Display name (3–30 chars; letters, numbers, _, ., - only) |
Yes | Profile identity; visible to your partner and, for parents, to your children's teen accounts |
| Email address | Yes | Authentication, transactional email, account deletion verification. A teen account also needs one; it may be a parent's address |
| Password (one-way hashed) | Yes* | Authentication (* not required if you sign in with Google or Apple) |
| Date of birth | No for adults (optional, collected during profile onboarding). A child's birth date is entered by the parent and is what decides when a teen account may be created. A caregiver must give their own date of birth when they redeem a code, and the account is refused if it makes them under 18 | Profile completeness; the age gate for teen accounts (§10.2); the adult gate for caregiver accounts (§10.13). Self-declared in every case — not a verification mechanism |
| Account role (Adult, Teen or Caregiver) | Set by the system | A teen account reaches only the teen surfaces and a caregiver account only the caregiver surfaces; every couple feature answers "forbidden" to both, enforced on the server. A caregiver account is an adult account in age only — it is not half of a couple and cannot become one while it holds a child |
| Country, city | No | Localization; visible to your partner only as part of your profile if they view it |
| Bio | No | Profile customization |
| Gender (male, female, other, or prefer not to say) | No (optional, in your personal details) | Profile completeness. Not shown to your partner; read by Refairly, the assistant, with your account in your own Refairly chat when it needs it (§4.7) |
| Profile picture (with auto-generated 64px thumb and 256px medium variants on AWS S3, server-side encrypted) | No | Profile display; visible to your partner once you connect, through private links that expire after a day. Max 10 MB; JPEG / PNG / WebP. The earlier picture is deleted from storage when you change or remove it, and every size of it when your account is permanently deleted |
Preferred reply language (ISO 639-1, default en) |
No | Drives the language AI replies are written in |
| Consent records | Created by your actions | For every consent you give (§2.3), for your acceptance of the legal documents at sign-up or in-app, and for the safety notice you confirm you have read when you sign up ("Before we go further": what Refairly is not, and who to call in an emergency): the scope, the version of the text you saw, the time, your IP address, your device's user-agent, and where it happened; for your own-health consent, which you give by adding a medicine, the time and that medicine instead of your IP address and device (§2.3). Kept as evidence of what you agreed to and when, including after your account is deleted (§8) |
2.3 Sensitive & special-category data, and the consents we ask for
Refairly's free-text surfaces invite users to write substantively about their relationship, their day, conflicts, their children and shared plans. Because of the nature of the product, users may foreseeably include content touching on health, family matters, religious or political views, sexual orientation, biometric or criminal-allegation-related matters, or other information that may constitute special-category data under GDPR Article 9 or sensitive personal data under KVKK Article 6.
The free-text surfaces with the highest sensitivity risk:
- Conflict descriptions (each partner's free-text describing the conflict)
- Conflict private session messages (each partner's 1:1 chat with the AI during mediation — the AI uses these to write a behavioral assessment that is kept for the life of the relationship)
- Conflict shared summary fields (
Felt,NeedKnow,TryNext— once approved, become visible to your partner) - Quick Repair messages (shared with the other person immediately)
- Joint onboarding answers (raw answers are private; only the AI-synthesized "About us" is shared)
- Growth Plan questionnaire answers (raw answers private; the AI-written reading and first goals are shared)
- Goals and their notes, photos and files (shared with the people who keep that list — §2.1)
- Daily Reflection answers (private to you; sent to AI only with your Reflection Insights consent)
- The couple's chat (shared with partner; read by the AI when either of you asks Refairly — §4.7)
- Your Refairly chat, and Refairly's notes about you (private to you; every message goes to the AI with the notes; you see and delete the notes in Settings → What Refairly remembers)
- Trip Plan wish (max 500 chars; seen by everyone on the plan)
- Trip Plan user-provided stay (label, neighborhood, optional latitude/longitude — seen by everyone on the plan; may identify a third party such as a friend whose home you're staying in)
- Trip Plan chat messages (shared; max 2000 chars per message; on a plan with no young person with her own account on it, every message goes to Anthropic with up to 20 recent messages; with one on it, only a message that writes @refairly, or one written while Refairly's question before a build is open, goes with them)
- Trip Plan lines of things to plan around (one per person, max 200 chars; may be health data or show a religion; sent to Anthropic for that trip's plan; seen only by the adult it is about and by the parents of a child it is about; deleted when the trip ends)
- Taste profile (only you see it; your trip plans are made from it)
- The "expecting a child" fact (health data; visible to your partner)
- A child's record and a parent's observations about the child (children's data; visible to the child's parents)
- A child's health list (allergies, intolerances, conditions; health data about a child; visible to the child's parents and to a caregiver they have given that child — §10.13; read by Refairly for a parent who asks — §4.7)
- The Day's medicine items and the records of doses given or not given (health data about a child; visible to the people §2.1 lists for the Day), and the alarms set on a medicine (seen in the app by the people who can see the item; on a lock screen they do not name the medicine — §2.10)
- A child's papers (files the parents add — health reports, school letters, photos; health data where a report is one, visible only to the people the parent chose per file — §10.14)
- Your own medicines and their records (health data about you; kept on your own-health consent, which adding your first medicine gives; seen by you and, in a couple (not co-parents), your partner — §2.12)
- The other items on your own day, such as an appointment's title (they do not rest on the own-health consent: in a couple (not co-parents) your partner sees them, and Refairly reads them like any other item, so keep health details out of them — §2.12)
- Mom or Dad's health list, medicines, papers and their records (health data about them; kept only once they have agreed in person — §10.15)
- A teen's check-in entries (a young person's own words about their wellbeing; her parents see the mood numbers she picks, and her words only while she shares them — §10.3)
- Family chat messages (visible to every member of the chat; read by the AI when a member asks Refairly — §4.7)
The consents we ask for in the app. Several surfaces require a one-time, explicit, recorded consent before first use. Each consent names a version of its text; if the text changes, we ask again. Each is recorded with the time, your IP address and your device's user-agent, and each can be withdrawn by writing to aoci@refairly.app, and in Settings where the app shows a switch for it: Reflection Insights, the fact that you are expecting ("Pregnancy information"), and a young person's themes in hers. The box for Refairly's AI features is asked for, recorded and taken back the same way, but it is a separate confirmation, not a consent (see its row). Your own-health consent works differently: nothing asks for it — you give it by adding a medicine to your own day yourself, beside one line that says what that means; it is recorded with the time and the medicine that gave it (no IP address or device); its words have a version too, and when they change, the next medicine you add gives it to the new ones; and you take it back by deleting your last medicine, or by writing to us (§2.12). The statement a son or daughter confirms before adding their mother or father is not a consent, so it is not in this table (§10.15):
| Consent | Who gives it | What it covers | Effect of withdrawing |
|---|---|---|---|
| Refairly's AI features | Each adult, once, in the app beside the legal documents — when signing up, or on the documents screen after signing in — by ticking its own box, separate from the box for the documents. This permission is required to use Refairly, which is built around its AI: the step goes on only once it is ticked. A caregiver is not asked, since a caregiver has no Refairly (§4.7) | Sending what you write, and what a feature needs, to our AI provider, Anthropic, in the USA or other countries, to make the answer: your Refairly chat, with the notes Refairly keeps about you; a chat's recent messages when you write @refairly or use "help me say it" there; every message you write in the chat of a trip plan with no young person with her own account on it, with the chat's recent messages; trips; the plan and "About us"; Reflection Insights; the Kitchen. This sending is necessary to provide those features, and that is its legal basis (performance of the contract, §3). The box is not a consent this sending rests on (GDPR Art. 6(1)(a)), nor an explicit consent under KVKK: it is a separate confirmation, apart from the documents, that you know what you write is sent to our AI provider. Asked as well as the Kitchen, Reflection Insights and children's-data consents, not instead of them | You can withdraw it at any time in Settings. The AI features then stop for you (a request already being processed may still finish): Refairly does not answer you or keep new notes about you (the notes it kept stay until you delete them), your @refairly and "help me say it" are not sent, you cannot write in the chat of a trip plan with no young person on it (every message there goes to Refairly), no trips, plan or "About us" are written from a request of yours, Reflection Insights stops and you cannot take a Kitchen step that goes to the AI. Since Refairly is built around its AI, it cannot be used without this permission: the next time you open the app, it asks you for it again, and you can log out or delete your account instead. Until then you can give it again in Settings or where an AI feature is used. Your records stay until you delete your account. What you have already shared with others — your taste profile, your trip wish and lines, your plan and "About us" answers, your messages in a chat and at the Kitchen — is still sent to Anthropic when someone who has given this permission asks for something that uses it; in the chat of a trip plan with no young person on it, that is every message someone else writes there. A young person is not asked this: her teen contract names Anthropic, and the safety check continues (§10.5) |
| Kitchen chat | Each adult | Sending Quick Repair conversations to our AI provider (Anthropic, in the USA or other countries) | You cannot start new Quick Repair threads |
| Kitchen table | Each adult | Sending Conflict descriptions and private sessions to our AI provider (Anthropic, in the USA or other countries) | You cannot start new Conflicts |
| Kitchen (young person) | The young person, on her own sheet — not a parent on her behalf | Sending what she writes in her Kitchen chat or at her table to the AI, and the safety check on it; her other parent seeing that a chat or a table exists, with whom, its status and a signed agreement | She cannot start or join a Kitchen chat or table; what exists stays readable |
| Reflection Insights | Each partner, optional | Distilling your private daily reflections into a private digest (§4.6) | The digest is deleted; reflections are still saved, still private |
| Expecting a child | Either partner, on the sheet shown before a "yes" is saved | Storing the fact that you are expecting, which your partner sees, and using it so the AI can support you both: it goes, with the couple's other facts, to our AI provider, Anthropic, in the USA or other countries, when the AI writes "About us", in a couple's Kitchen chat or table, when Refairly answers in the couple's chat or in either partner's own Refairly chat, and in trip plans for the two of you (§4.1, §4.7) | In Settings ("Pregnancy information") or by writing to us: the fact is removed for both partners, whichever of you withdraws, from every relationship of the one who withdraws that holds it, ended ones included, and is no longer sent; the "About us" written with it is cleared and written again without it the next time either of you opens it, and Refairly's cards that quote it are emptied. Refairly's earlier replies stay in the chat's recent history, which goes to Anthropic with each new message |
| Children's data | Each parent, before adding the first child | Storing what parents keep about a child — the record and its card, the health list, observations, the Day with its medicine records, the child's goals with their notes, photos and files, and the papers they add, health reports included (§10.14) — and Refairly reading them and suggesting changes the parent taps (§4.7), for which they are sent to our AI provider, Anthropic, in the USA or other countries. The consent text names all of it, and says plainly that the health list, medicine records and health reports are health data | You cannot add new records, observations, papers or files on a child's goals; what exists stays for the other parent until the record is deleted. In its chats, Refairly reads nothing new about the child for this parent — the record and its card, the health list, observations, the Day with its medicine records, goals, papers, files and pocket money — and suggests no changes about the child; its earlier replies stay in the chat's recent history, which goes to Anthropic with each new message. The household plan's next reading is made without the children's names, ages and notes |
| Your own health | Each adult, by adding a medicine to their own day themselves: in the day's editor, where one line beside Medicine says that the medicines they add are kept for their reminders, who else sees them (when anyone does), that Refairly reads them through Anthropic and that deleting one removes it for good, or by tapping a card from Refairly that adds one. Adding the first one is the consent; nothing else asks for it. Only the person whose day it is gives it: a partner cannot give it for them, and cannot add a medicine to their day until it stands. Earlier versions of the app ask for it on a sheet before the first medicine | Keeping the medicines on your own day — their doses and times, and records of taken or not taken — with their reminders and alarms; your partner in a couple, now or one you connect with later, seeing them with their past, and adding, changing and ticking them; and Refairly reading them for you or your partner, for which they are sent to our AI provider, Anthropic, in the USA or other countries. These are health data. Only medicines rest on this consent: any other item on your day, such as an appointment's title, is seen by your partner and read by Refairly like any other item, so keep health details out of it (§2.12). The line in the editor is short: this row and §2.12 say in full what adding a medicine agrees to | Deleting your last medicine yourself takes it back — your partner deleting it does not — and erases besides only the medicines removed from your day earlier and kept out of sight (items that were once medicines included), with their records, and the past as a medicine of an item changed into another kind (each medicine you or your partner delete now goes for good anyway, with its records, reminders and alarms, and Refairly's cards about it are emptied); your partner cannot add a medicine to your day until you add one again. Taken back by writing to us, or with the switch "Medicines on your day" in Settings on earlier versions of the app, it deletes every medicine on your day at once, earlier removed ones included, with their records, reminders and alarms, and the past as a medicine of an item changed into another kind, and Refairly's cards about them are emptied. An item changed out of a medicine stays as it is now (§2.12). Either way Refairly reads nothing more about your medicines; its earlier replies stay in the chat's recent history, which goes to Anthropic with each new message |
| Mom or Dad's own answer: their health (in person) | Mom or Dad themselves — not their son or daughter on their behalf — in person, after reading or hearing the words the app shows for them. Their son or daughter records it; the record keeps who recorded it, when and which words | Keeping their health list, medicines and papers, with their records and reminders | Their son or daughter records that they changed their mind (or they write to us). Their health list, medicines and papers — the files on their goal items included — are deleted with their records, reminders and alarms, and so is the past as a medicine of one of their medicines changed into another kind (§10.15); Refairly's cards about them are emptied |
| Mom or Dad's own answer: Refairly's assistant (in person) | The same, as a separate answer on the same page | Refairly reading what is kept about them, for which what it needs is sent to our AI provider, Anthropic, in the USA or other countries | Recorded the same way. Refairly reads nothing more of their record; its earlier replies stay in the chat's recent history, which goes to Anthropic with each new message |
| Teen wellbeing | The young person, at claim time — not a parent on their behalf | Turning their check-in entries into themes for parents; AI replies in chats they are in; her goals with their notes and files, which her parents see; and her own Refairly's cards | Themes already made are deleted; entries stay saved; the AI no longer replies in their chats, her Refairly makes no cards, and a parent's Refairly does not read chats she is in. Her goals, notes and files stay visible to her parents (that rests on the parent's children's-data consent). The safety check continues (§10.5) |
| Legal documents | Everyone, at sign-up and again when the documents change materially | Acceptance of the Terms, and confirmation that you have read this Policy and the KVKK notice | You cannot use the Service until you accept the current version; you can log out or delete your account |
Our position on sensitive data:
- Our structured questionnaires are designed to elicit feelings, needs, and proposed outcomes — not to invite sensitive disclosures.
- In the couple features the Service does not perform server-side keyword detection, abuse-content filtering, or self-harm signal extraction. Safety-conscious AI behavior there is enforced only by AI system prompts and is not a guaranteed safeguard. The family surfaces carry the automated safety check described in §10.5.
- Where you nonetheless include sensitive data, we process it solely to (a) provide the feature you invoked, (b) maintain the integrity and safety of the Service, and (c) comply with legal obligations or defend legal claims where necessary.
- You must not include sensitive data about another person (a partner, a child, a family member, a third party named in a conflict) unless strictly necessary to describe the situation and you have a lawful basis to disclose it. A parent may record about their own child what the child's care needs — observations written in care, the health list, medicine on the Day, and health reports as papers; a parent must not record other people's data about the child, or anything the child's care does not need. A son or daughter records health details about their mother or father only once they have agreed in person (§10.15).
- We may restrict, redact, remove, or decline to process submissions that contain obviously excessive, gratuitous, unlawful, or harassing sensitive data.
- We do not use sensitive data for advertising or marketing, and we do not sell it. Some features analyse what you write to produce that feature's result (for example the Kitchen's private assessment, §4.3); we use it for nothing else beyond safety review, abuse prevention, and related legal/compliance purposes where applicable. When you give a consent, our app analytics, if you turned it on (§5.4), records only that a consent was given, not which one. Under the commercial terms we use, our AI provider does not train its models on your data (§5.2).
- Sensitive content is held in PostgreSQL on AWS RDS (encrypted at rest via AWS-managed encryption) and is subject to the retention, purge, deletion, and legal-hold rules described in this Privacy Policy.
- Withdrawal does not affect processing already carried out before withdrawal, nor any retention required by law or reasonably necessary for the establishment, exercise, or defense of legal claims.
- We strongly advise you to include only information directly relevant to what you are trying to accomplish. Do not submit government ID numbers, financial account numbers, passwords, access credentials, biometric identifiers, or detailed medical records unless strictly required.
2.4 Couple-shared content (what your partner sees)
The complete per-feature breakdown is in §2.1. The most consequential crossings:
- Family pulse — your one-word mood is visible from the start to your partner (a co-parent too), and between a parent and each teen they parent who has her own account; each of them is notified when it changes. Only the word, not anything else you wrote. You can stop sharing with any of them at any time, except when only one other person shares your household — the two of you, or a parent on their own and one teen — who then always sees it.
- Relationship facts — visible to both, including the "expecting a child" fact.
- The couple's chat, Trip and Quick Repair shared chat threads — everyone in them sees every user message and every AI message in real time, including Refairly's cards in the couple's chat; only the one who asked can tap a card.
- Growth Plan reading + household goals — both partners read and edit the household's goals, with their notes, photos and files; a young person with her own account reads the ones not marked Parents only (§2.1). When a parent on their own connects a partner, their plan joins the couple: the partner sees its goals and, if it was woven from that parent's answers alone, its reading until both have answered again. A reading woven with an earlier partner is cleared first, so a new partner does not read it.
- Each partner's own goal items — while the couple is active, each of you reads, adds to, changes and deletes the other's, with their notes, photos and files; a young person at home with you reads the ones not marked Parents only (§2.1). Items an adult kept from an earlier couple are seen by the new partner too.
- Trip Plan content, its choices and picks, and any user-provided stay — everyone on the plan, and the parents of a child on it, see all of this.
- Taste profile — only its owner (a child's: her parents).
- The Kitchen table's proposals + agreement + signatures — shared.
- Conflict shared summary (
Felt,NeedKnow,TryNext) — becomes shared after you approve your own row. - Joint onboarding — only the AI-synthesized "About us" is shared; your raw answers are not.
- Child records, the children's goals and Day, the household's Home day, expenses and family chats — shared with your partner, the child's other parent (§10.3). The alarms on goals and Day items are seen by everyone who sees the goal or the item, with whom each rings and how each person answered (§2.10). A top-up you give a child as pocket money also appears in the household's money.
- Your own day — in a couple (not co-parents), your partner sees all of it, your medicines included, and may add, change and tick your items; your partner can add a medicine only while your own-health consent stands, which your own first medicine gives (§2.12).
- Mom or Dad — a mother or father one of you looks after is seen by the other partner through the partner link while the couple lasts, notes and items marked Parents only included (§10.15).
- A pet — in a couple (not co-parents) both of you keep it, with its card, day, notes and papers (§2.1).
- Subscription state — your partner can see a subscription is active and which partner is the payer (
payerUserId).
By connecting with a partner you acknowledge the above sharing is inherent and necessary to the features described and is performed on the legal basis of performance of the contract for the relevant feature, not on separate per-share consent.
2.5 Technical & device data
- Device push tokens — FCM token (Android and most iOS pushes) + APNs VoIP token (iOS Buzz Call invitations only)
- Per-device metadata — platform, app version, OS version, device id and name (as supplied by the mobile app)
- IP address — logged by our web server for security, rate-limiting, and abuse prevention; recorded with each consent you give and with each safety report you file
- Authentication tokens — access tokens stored on your device via platform-secure storage; one refresh token per signed-in device stored server-side as a hash, with a short device label, so that signing in on a second device does not sign the first one out
- Password reset codes + account-deletion verification codes — temporarily stored on-device and server-side with short-lived expiry (15 minutes for the deletion code)
- Notification data — see
NotificationHistoryEntry,NotificationPreferences,DeviceTokenin §2.1 - Audit metadata —
CreatedAt/UpdatedAttimestamps and user identifiers recorded on rows for security auditing - AI usage telemetry —
TokenUsageLogrecords token counts, estimated USD cost, AI provider name, AI model name, request-type tag (e.g.,"conflict_synthesis","tripPlan_generation") per AI call. No prompt content, no AI response content, no user-typed text is included in these logs. - Error monitoring — when our server hits an error, an event with the error, the request path and, where our code attaches it, your user identifier may be sent to Sentry (hosted in Germany, §5.1). Our logs are designed not to contain message bodies or AI prompts, and email addresses are masked before logging.
2.6 Subscription & billing data
A subscription belongs to the account that buys it. In a couple, one partner's subscription covers both; if both subscribe, the two add up to one pool — each kept on its subscriber's UserSubscription row, spending drawing from the one closest to renewal first, the higher plan setting the features. A user without a partner can subscribe for their own account. The family features, including a teen's AI usage, draw on the same pool — the couple's, or a solo parent's own (§12). A teen account cannot buy anything and is not charged.
The welcome week. Every new household starts with seven days in which everything is open, the AI included, on one small allowance the whole household shares; no plan is chosen and nothing is bought to start. The week begins when its first adult account is made, a partner who connects shares the days left, and a household that has ever had a plan has no welcome week (§12).
We persist:
- Tier (None / Weekly / Pro / Max), monthly token limit (computed from tier), tokens used this period (computed), bonus tokens (from token packs; they do not expire), current billing-period window (computed), and when the welcome week ends
- The store's identifiers for your purchase (transaction ids, which store, the product and when it expires), all from Apple / Google, and when we last checked its status with the store
- A record of each store notification we processed (its id and type)
- A history of changes to your plan (for example a renewal, an expiry, a refund or bonus tokens added), with the old and new values and where the change came from
- A pseudonymous account token per store, used to link a store purchase to your account; deleted with the account
We do not collect, store, or process credit-card numbers, bank details, or payment credentials. Apple and Google handle all payment. We receive only transaction identifiers and subscription status from their APIs / webhooks.
2.7 Invitation data
When you invite someone to connect with you on Refairly:
- The invitation is stored as a
ConnectionRequest(sender id, receiver id, optional note, status, timestamps). Invitations require both parties to already have an account; we do not support inviting a non-user via email or SMS. - If you misuse the invitation system to harass, this is a violation of our Terms of Service.
2.8 Data we do NOT collect
We don't collect your contacts, the phone number of an account holder, background location, advertising identifiers or call audio, and we don't show ads. Buzz Call intensity samples pass through our real-time hub and are not stored (§2.11). Payment-card details are handled by Apple and Google. We collect children's data only as §10 describes, and data about a mother or father the family looks after only as §10.15 describes; no child under 13 has an account.
2.9 Device permissions
The mobile app may request the following device-level permissions. Each is requested in context, requires your explicit consent, and can be revoked at any time through your device settings. None of these permissions enables background data collection or tracking.
- Camera — when you choose to take a profile photo, when you scan your partner's invite code to connect, and when a young person or a caregiver scans the code shown on a parent's or keeper's phone. A scanned code is used once and not stored as an image.
- Photo library and files — only when you choose a profile picture, attach a receipt photo to an expense, add a paper about a child, Mom or Dad, or a pet, or add a photo or file to a goal, through the system picker. We receive only the file you selected.
- Notifications — for push notifications. Granular per-type preferences are available in app settings; other notification types respect your master push switch and quiet hours.
- Alarms — so an alarm you accepted (§2.10) rings on time like the clock app. On Android: notifications, exact alarms, and showing an alarm full screen over the lock screen, which Android 14 and later ask you to allow in the phone's settings. On iPhone with iOS 26 or later: the alarm permission (AlarmKit); on an earlier iPhone an alarm rings as a time-sensitive notification. The app explains these the first time an alarm is set to ring your phone and opens the setting when you tap. A "no" keeps the alarm saved; without exact alarms it can come a few minutes late, without the full screen it shows as a notification, and on Android or an iPhone before iOS 26, without notifications it cannot show.
- Haptic engine — for Buzz and Buzz Calls. No data is collected; the engine renders patterns locally.
- Microphone — for the voice in a Buzz Call (§2.11). On Android the app asks when you place a call, and when a call you answered connects while the app is open; on iPhone the system asks the first time a call's audio starts. The app uses the microphone only during a call, once it is answered, and you can mute it during the call. If you say no, the call goes ahead with vibration only.
We do not read your contacts or your device's location — your city and country are what you type, and approximate stay coordinates in the Trip Planner are what you enter (§2.1) — and we use the microphone only as above.
2.10 Alarms
What an alarm is. An alarm belongs to one Day item or one goal and rings at a rule: once, on a date and at a time; at a time on chosen weekdays; a number of days before a goal's date; or at a Day item's own time, following the item when its time moves. It can have a heads-up before it (minutes or hours before on a Day item, days before on a goal) and a short label. An item or a goal holds at most 20 alarms. An alarm rings at its time on each person's own clock: each phone tells our servers its time zone when you sign in and when you come back to the app, so an alarm set for 08:00 rings at 08:00 on every phone it is for, wherever that phone is. Until a phone has told us, an alarm rings there by the clock of the day it belongs to: on a child's, Mom or Dad's or a pet's item or goal, that person's clock; on the Home day, the Home day's; on an adult's own item or goal, the clock of that adult's phone; and on a household goal, or where no clock is known yet, UTC. An alarm on a Day item does not ring on a date the item is skipped, paused or stopped, or once someone has recorded that date's item (ticked it, or recorded a dose as given or not given); an alarm on a goal stops when the goal is done.
Who it rings, and asking first. An alarm can ring only people who can see what it is on: the child's parents; the young person herself, on her own goals and the Day items she sees; a caregiver who sees that Day item or reads that child's goal (§10.3, §10.13); on the household's goals and the Home day, the partners or the parent on their own; on a partner's own goals, both partners; and on these three, a young person at home with them, where the item is not Parents only (§2.1); on your own day, you and your partner; on Mom or Dad's day and goals, their son or daughter, that person's partner and their caregiver; on a pet's day, its keepers, its caregiver and a young person at home with it. Whoever may change the item or goal can name any of them; anyone else who can see it can set an alarm for themselves alone, or add themselves to one that is there. An alarm for yourself is set at once. For anyone else it is a request: they are told who asks, what it is for and when, and they accept — choosing their own heads-up — or decline. It does not ring them until they accept, and they can change their answer later. Adding someone to an alarm asks them; changing the time of an alarm someone accepted tells them and keeps their yes. A request nobody answers expires seven days after it was asked, and a request for an alarm with a single moment also expires once that moment has passed. Anyone who can change both the alarm and the item or goal it is on can ask again — directly, or by moving the alarm to a later moment — which starts the seven days over. Everyone an alarm is for can turn it off for themselves and set their own heads-up without changing it for the others. Whoever set it can change or remove it, and so can the child's parents on her things, either partner on the household's goals and the Home day, both partners on a partner's own goals and on a partner's own day, Mom or Dad's son or daughter on their things, and a pet's keepers on its day. A paused young person sets no alarm, and her phone is given none to ring, until her account is back on.
Where it rings. Our servers keep each alarm as a rule. Each person's phone reads its own alarms for the next two weeks and schedules them in the phone's own alarm and notification system, so they ring without a connection at that moment: on Android, a full-screen alarm over the lock screen; on an iPhone with iOS 26 or later, an iOS alarm once you allow it; otherwise a time-sensitive notification (§2.9). For each alarm it has scheduled, the phone holds what it is for, the child's name and the time. Whether and when it rings depends on that phone — its permissions, its settings and its battery; it may not ring if the phone is off, silent or restricted. For a medicine, keep a second reminder of your own. Logging out of the app on a phone removes the alarms from it. "Done" on a ringing alarm opens the item or the goal; it records nothing — a dose is recorded only by a person who records it on the Day. Snooze works on the phone alone. Our servers do not record whether an alarm rang, was snoozed or was stopped.
What the notifications say. A request ("{name} asks to alarm you") and a change of time ("{name} changed an alarm") are push notifications to the people concerned, also kept in their in-app inbox (§8), with what the alarm is for and when it next rings. On a lock screen an alarm on a medicine does not name the medicine: the alarm, its heads-up and these notifications read "Time for {name}'s medicine" ("Time for your medicine" to the young person herself, and to an adult on their own day), and the medicine's name and the alarm's label show in the app once the phone is unlocked. Any other alarm shows its label or the title of its item or goal, and may show the child's name. When an alarm or what it is on changes, the phones of everyone it is for receive a silent push that carries nothing but "alarms changed", and fetch their alarms again; a phone that has not fetched yet rings what it already holds.
What we keep, and for how long. Alarm and AlarmPerson as in §2.1. A one-time alarm is deleted 30 days after its date. An alarm goes when the Day item, goal or child's record it is on is removed, and when nobody is left on it; a person who can no longer see what it is on comes off it. An alarm on a goal that is done, or on a Day item that has ended, rings nothing and stays until it, or what it is on, is removed. When you ask to delete your account, a request you sent that nobody has answered is withdrawn; at erasure your place on every alarm goes, and an alarm you set stays for the people on it, without your name (§9.2). Your alarms are in your data export (§9.1).
2.11 Voice in Buzz Calls
What a call carries. A Buzz Call carries the live vibration and, when the microphone is allowed, your voices: the two people on the call hear each other. If either of you does not allow the microphone, or the voice cannot connect, the call goes on with vibration only.
When the microphone is asked for. On Android, the app asks when you place a call, and when a call you answered connects while the app is open. On iPhone, the system asks the first time a call's audio starts. You can change your answer in your phone's settings at any time. The app uses the microphone only during a call, once it is answered, and you can mute it during the call.
Where the audio goes. The audio goes directly between the two phones over WebRTC, encrypted in transit, and does not pass through our servers. We do not record calls, and we do not store call audio or send it to any AI.
What our server sees. To connect the audio, the two phones exchange set-up messages through our real-time hub: a description of the audio session and the network addresses each phone can be reached at, which include its IP addresses. The hub passes them to the other phone, holds them in memory only while the call lasts, and does not store them. The call record (BuzzCall) keeps who called whom, in which chat, whether it was a voice call, when it started, was answered and ended, how it ended and how long it lasted — no audio and no network addresses. A finished call's record is deleted 365 days after the call started, or earlier with your account (§8). Where you turned analytics on (it is never on for a teen account), the app also sends Firebase Analytics a short connection-quality note for a voice call — the kind of connection and the delay, without any audio (§5.4).
Google's STUN servers. To find a direct path, each phone asks Google's public STUN servers for its own public address. Those servers see the phone's IP address; no call audio passes through them (§5.4).
No relay server today. Some networks do not allow a direct connection between two phones. We do not run a relay (TURN) server today, so on such networks the voice does not connect and the call goes on with vibration only. If we add a relay, call audio on those networks would pass through it, still encrypted; we will name it in §5.4 and describe it in this section before it is used.
2.12 Your own day and your own medicines
What it is. Every adult has a day of their own from the start — in a couple, on their own, or as a co-parent; a caregiver's account has none. It is built like a child's day: items at a time or for the whole day, which can repeat, and a tick when each is done. In a couple (not co-parents), you and your partner can both add, change and tick items on it, and every change shows who made it. When your partner adds, changes or takes off something that falls today, you are told; a tick tells nobody. Your day stays yours after a couple ends, and after the two of you become co-parents.
Who sees it. You and, in a couple (not co-parents), your partner, all of it; otherwise only you: there is no item one of you keeps from the other. A partner you connect with later sees your day, its past included. Your children, a young person with her own account, and anyone who helps look after someone in your family do not see it. When the couple ends, or the two of you become co-parents, each of you keeps your own day and goal items, and the other's access ends at once.
Medicines and health details. Medicines on your day are health data, kept on your own-health consent (§2.3). Only medicines rest on that consent. Any other item on your day, such as an appointment's title, is seen by your partner in a couple and read by Refairly like any other item, whether or not your consent stands. So please keep health details out of those items — what an appointment is for, a diagnosis, a symptom — and so should your partner when they add or change one. Adding a medicine to your own day yourself is that consent: beside Medicine, the day's editor says in one line that the medicines you add are kept for your reminders, who else sees them (when anyone does), that Refairly reads them through Anthropic and that deleting one removes it for good, and saving your first medicine records your consent, with the time and that medicine; tapping a card from Refairly that adds one does the same. Nothing else asks for it, and it covers everything §2.3 lists: your partner, now or one you connect with later, seeing your medicines with their past and adding, changing and ticking them, and Refairly reading them for you or your partner, for which they are sent to Anthropic, in the USA or other countries. You give it yourself; your partner cannot give it for you, and can add a medicine to your day only while it stands. A medicine's reminder comes to you; your partner is reminded too only after choosing "Ring me too". A reminder on a locked screen says only "Time for your medicine", and a notice that your day changed says only that a medicine changed. A medicine deleted from your day — by you or your partner — is deleted for good, with its changes, its records of taken or not taken, its reminders and its alarms, and Refairly's cards about it are emptied. Deleting your last medicine yourself also takes your consent back — your partner deleting it does not — and with it the medicines removed from your day earlier (items that were once medicines included), which had been kept out of sight with their records, are erased for good, and so is the past of an item changed out of a medicine (below); nothing else goes with it, and adding a medicine gives the consent again. Taking it back by writing to us deletes every medicine on your day at once, earlier removed ones included, with their records, reminders and alarms; so does turning off "Medicines on your day" in Settings on earlier versions of the app, which ask for this consent on a sheet before the first medicine. You can also change an item on your day into a medicine — yourself, which gives your consent as adding one does, or, while your consent stands, your partner — and change a medicine into another kind of item, such as a walk or a meal. A medicine changed into another kind keeps its past — its earlier names and notes and its records of taken or not taken — and your consent stays as it was, even if it was your last medicine. From then on it is not a medicine, and Refairly reads it as any other item on your day. Its past as a medicine goes with your consent: taking your consent back, in any of the ways above, erases that past for good and leaves the item as it is now. If you delete such an item yourself, it is deleted for good with its past; if your partner deletes it, it is erased with its past 30 days later, as any removed item is (§8).
Your clock. Your own day follows your phone's time zone: when you travel, a medicine set for 20:00 reminds you at 20:00 where you are.
What we keep. As §8 says. Your day is in your data export (§9.1).
3. Legal Bases for Processing
| Processing activity | Legal basis (GDPR) | Notes |
|---|---|---|
| Account creation and authentication | Performance of contract (Art. 6(1)(b)) | Necessary to provide the Service |
| Each couple feature in §2.1 (Daily, Growth, Goals, Trip, Conflict, Quick Repair, Buzz, Buzz Calls), and the household's Home day | Performance of contract (Art. 6(1)(b)) | Core service delivery |
| AI processing of your content (per-feature) | Performance of contract (Art. 6(1)(b)) | Necessary to provide Refairly's features, which are built around the AI. The AI permission (§2.3) is asked in addition to this basis, in a box of its own, and is required to use Refairly. Includes cross-border transfer to Anthropic — see §6 |
| Processing of special-category data in free-text submissions, and of the Kitchen surfaces | Explicit consent (Art. 9(2)(a)) | See §2.3 — consent may be withdrawn at any time |
| The "expecting a child" fact | Explicit consent (Art. 9(2)(a)) | Health data |
| Trip Planner: the line of things to plan around (food rules, allergies; may reveal health or religion) | Explicit consent (Art. 9(2)(a)): your own line, given when you write it or say it in the plan's chat; a child's, the parent's children's-data consent | Kept for that trip only: deleted when the plan ends or 7 days after its last day; clearing the line withdraws it |
| Reflection Insights | Explicit consent (Art. 9(2)(a)) | Optional; §4.6 |
| The relationship check and its trend label (§2.1), and measuring in aggregate whether the Service helps couples | Performance of contract (Art. 6(1)(b)) for the check and its label; legitimate interest (Art. 6(1)(f)) for the aggregate measurement | Answers are not shown to the partner or read by Refairly; the internal dashboard shows counts, averages and how scores move, not anyone's answers |
| Child records and the child's card, parents' observations, the health list, the Day and its medicine records, a child's goals with their notes, photos and files, child-scoped expenses and pocket money | Consent of the parent (Art. 6(1)(a)); explicit consent for the health list, medicine records, health reports and wellbeing observations (Art. 9(2)(a)) | Children's data merits specific protection (Recital 38). Parents represent that they hold parental authority (§10.2) |
| Your own day, apart from medicines, and, in a couple, sharing it with your partner | Performance of contract (Art. 6(1)(b)) | Every adult has one; sharing with your partner is the feature. Its other items do not rest on the own-health consent, so health details should be kept out of them (§2.12) |
| Your own medicines, their records, reminders and alarms, and Refairly reading them | Your explicit consent (Art. 9(2)(a)), given by adding your own first medicine yourself, beside the line that says what that means (§2.3, §2.12) | Deleting your last medicine takes it back. Only medicines rest on this consent (§2.12). Includes cross-border transfer to Anthropic — see §6 |
| Mom or Dad's record apart from their health list, medicines and papers — name, birth date, whose parent they are, card, their day without medicines, goal items and notes — and sharing it with their son or daughter's partner and their caregiver | Legitimate interests (Art. 6(1)(f)): our interest in providing the feature their family uses, and their family's in looking after them | Balanced by keeping no health list, medicine or paper and sending nothing of their record to the AI before they agree, by their son or daughter telling them what is kept, and by asking their son or daughter to leave health details out of the rest. They can object at any time, through their son or daughter or by writing to us; an objection deletes the record |
| Their health list, medicines, papers and their records | Their explicit consent (Art. 9(2)(a)), given in person and recorded by their son or daughter | — |
| Refairly reading what is kept about them | Their explicit consent (Art. 9(2)(a)), a separate answer, given in person | Includes cross-border transfer to Anthropic — see §6 |
| The record of their answers | Legal obligation to demonstrate consent (Art. 6(1)(c), Art. 7(1)) and legitimate interest (Art. 6(1)(f)) | Who recorded each answer, when, which words; no health data |
| A pet's record | Performance of contract (Art. 6(1)(b)) | Kept as part of its keepers' account; contact details typed under "who to call" are a third person's data, kept for the keepers' use |
| Refairly, the assistant: reading what the person who asks may see, applying the cards that person taps, and keeping the record of each turn in a person's own Refairly chat (§4.7) | Performance of contract (Art. 6(1)(b)); for the special-category data it reads — a child's health list, medicine records and health reports — the parent's explicit consent given with the children's-data consent (Art. 9(2)(a)); for an adult's own medicines, that adult's own-health consent (Art. 9(2)(a); §2.12); for a young person's own data, her consent (Art. 6(1)(a); Art. 9(2)(a) where it holds health data) | Includes cross-border transfer to Anthropic — see §6. A card changes nothing until the asker taps it |
| Refairly's notes about an adult: kept from their own Refairly chat, and read into its later turns there (§4.7) | Performance of contract (Art. 6(1)(b)); for a detail about their own health, religion or beliefs, sex life or sexual orientation, ethnic origin, political views or union membership that the person asked Refairly to remember, their explicit consent, given by that request (Art. 9(2)(a)) | Adults only: none about a young person or a caregiver. Refairly is told never to keep such a detail about anyone else. What a note may hold is set by instructions to the model, not by a filter (§4.7). Each note can be seen and deleted in Settings → What Refairly remembers; all are deleted with the account. Includes cross-border transfer to Anthropic — see §6 |
| A parent's pocket-money top-up written into the household's money | Performance of contract (Art. 6(1)(b)) | The same amount, date and note in both places (§2.1) |
| Alarms (§2.10): keeping them, asking the people named, and telling their phones when an alarm changes | Performance of contract (Art. 6(1)(b)); an alarm on a child's item or goal is part of the child's record and rests with it on the parent's consent (row above; Art. 9(2)(a) for an alarm on a medicine) | An alarm rings someone else only after they accept it. On a lock screen an alarm on a medicine does not name it |
| A teen's account, check-in entries and themes | The young person's explicit consent (Art. 9(2)(a)), given in a flow that a parent initiates and that constitutes the parental authorization Art. 8 requires where it applies | The parent-initiated claim is our parental-consent mechanism in every country, whatever that country's age of digital consent (§10.9) |
| The automated safety check in family surfaces (§10.5) | The young person's consent given at claim time, which expressly includes it; where that consent is withdrawn, protection of vital interests (Art. 6(1)(d) and Art. 9(2)(c)) | Limited to showing help resources, pausing her Kitchen chat or table and keeping what she wrote there for her alone, and, for school and outside-life signals, asking a parent to check in. Not used for any other purpose |
| Sharing content between connected partners per §2.1 | Performance of contract (Art. 6(1)(b)) | Core feature; users are informed before connecting and before adding a child |
| Transactional emails (registration welcome, email confirmation, password reset, account-deletion verification code) | Performance of contract (Art. 6(1)(b)) | Necessary for account security |
| Marketing or notification-mirror emails | Not sent | Only the transactional emails above are sent |
| Push notifications | Consent (Art. 6(1)(a)) via device permission + in-app preferences | Master switch + quiet hours + per-type opt-out where offered |
| In-app purchase verification, subscription / token-pack management, and a household's welcome week | Performance of contract (Art. 6(1)(b)) | Receipt verification + Apple Server Notifications V2 + Google Pub/Sub RTDN |
| Rate-limiting, fraud prevention, abuse detection, safety reports | Legitimate interest (Art. 6(1)(f)) | IP logging, in-process rate-limit counters, idempotency-key tracking, human review of reports |
| Recording consents and legal-document acceptance with IP and device | Legal obligation to demonstrate consent (Art. 7(1)) and legitimate interest (Art. 6(1)(f)) | |
| App analytics (Firebase Analytics) | Consent (Art. 6(1)(a)), which also covers the analytics SDK keeping and reading an identifier on your device (Art. 5(3) of the ePrivacy Directive and the national laws that apply it) | Off until you turn it on — when you sign up, on the documents screen or in Settings — and you can turn it off in Settings at any time. Never on for a teen account |
| Crash reporting (Firebase Crashlytics) and server error monitoring (Sentry) | Legitimate interest (Art. 6(1)(f)) | Bug diagnosis |
| Server logs and IP addresses | Legitimate interest (Art. 6(1)(f)) | Security monitoring, incident investigation |
| Checking for app updates through Expo's update service (§5.4): the phone's IP address, its platform, the app's version and an installation identifier | Legitimate interest (Art. 6(1)(f)) | To keep the app working and up to date. Expo receives no account details and nothing you write. Includes cross-border transfer to Expo — see §6 |
| Serving the website, refairly.app, from AWS S3 and CloudFront (§5.1): your IP address and the page you open | Legitimate interest (Art. 6(1)(f)) | To show you the pages you open, these documents and the account-deletion page included. We keep no access logs for the website |
| Defense of legal claims | Legitimate interest (Art. 6(1)(f)) | Retention of relevant records (signatures, agreements, audit trails, consent records) |
KVKK (Turkey): Under KVKK m.5 our processing is based on necessity for performing a contract to which the data subject is party (m.5/2-c), legitimate interests of the data controller where fundamental rights are not harmed (m.5/2-f), explicit consent for children's data, for the Kitchen surfaces and for app analytics (m.5/1), and explicit consent for sensitive personal data under m.6/3-a. For the safety check where a young person has withdrawn consent, we rely on the protection of life or physical integrity (m.5/2-b; for special-category data, m.6/3-c). For a mother or father the family looks after, we rely on m.5/2-f, our legitimate interest in providing the feature the family uses, for their record without health data; on their explicit consent, given in person, for their health data and for Refairly reading their record (m.6/3-a); and on m.5/2-e for the record of their answers. For an adult's own medicines, we rely on that adult's explicit consent (m.6/3-a), which they give by adding their own first medicine (§2.12); the other items on their day rest on m.5/2-c, and we ask that health details be kept out of them. For Refairly's notes about an adult we rely on m.5/2-c, and on that adult's explicit consent (m.6/3-a) for a detail about their own health, religion or beliefs, sex life or sexual orientation, ethnic origin, political views or union membership that they ask it to remember. Specific bases per category are itemized in the KVKK Aydınlatma Metni.
Canada / Quebec: For users in Quebec, the consents in §2.3 are the express consents Law 25 requires for sensitive information and for a minor under 14 (§10.9), and this Policy is the notice the Act requires at collection.
We do not sell your personal data. Monetization is through couple-tier subscriptions and consumable token packs purchased via Apple / Google. We do not display third-party advertisements.
4. AI Processing & Automated Decisions
4.1 Which AI provider runs which surface
Refairly currently uses Claude, made by Anthropic, for all AI surfaces. We do not currently use OpenAI, Google Gemini, or any other LLM provider for AI inference. Which Claude model a feature uses may change as models are updated; the recipient (Anthropic) and the purposes do not change.
| Surface | Notable payload |
|---|---|
| Joint-onboarding "About us" synthesis | Both partners' 5 raw answers |
| The couple's chat (only a message that tags @refairly) | The recent conversation and what both partners may see (§4.7). Output: a reply and up to three cards that only the one who asked can tap. Messages without the tag reach the AI only when a partner's own Refairly reads the chat (below, §4.7) |
| Reflection Insights digest (opt-in, §4.6) | Your own recent reflections since the last digest |
| Growth Plan questionnaire synthesis | Both partners' 5 raw answers + relationship "About us"; for a household, also each child's name, age and the note on the record, while both parents have given the children's-data consent (§2.3); output is the 7-section reading + the household's first goals. For a parent on their own: their own 5 answers and, while they have given that consent, each child's name, age and note. |
| Trip Planner — clarifier (pre-generation gate) | The trip's answers + who is coming (names, ages, lines of things to plan around) + "About us" when it is just the couple |
| Trip Planner — first generation (background worker) | Trip details + the names, ages, taste profiles and lines of the people coming + the candidate places |
| Trip Planner — chat (every message on a plan with no young person with her own account on it; with one on it, only when someone writes @refairly, or while its question before a build is open) | The current plan + up to 20 recent messages and the plan's last few events, with a short note after a person's message of what Refairly's plan tools did for it, or that its turn failed + who is coming and their lines. Output: a reply, and a change to the plan when someone asks for one. A rebuild it runs also carries the last 10 messages people wrote in the chat — on a plan with a young person on it, only those not written to Refairly |
| Trip Planner — per-place insight (Tell me more, on a stop's place or any of its choices) | Trip details + who is coming + their lines without names + the one place |
| Conflict — private session opening | Both partners' descriptions + your UserProfile.ProfileData + the household context (for a couple: "About us" and the couple's facts, the name and age of a child the table is about, and, for a table about money, this month's totals from the household's money; for a parent and a young person: the family block) + your Reflection Insights digest if any (a couple's table only) + the Kitchen's memory of up to five earlier closed Kitchen chats and tables between the same two people: their topics, how they ended, any agreement and whether it held and, for you alone, the AI's private assessment of you from each and your own check-in note, never the other person's |
| Conflict — private session follow-up | Same as opening + your full private-session transcript so far + coverage state |
| Conflict — shared-summary draft | Your full private-session transcript |
| Conflict — private suggestions ("Express Clearly / Keep In Mind / Try Differently") | Your full private-session transcript |
| Conflict — complaint assessment (per partner) | Your description + your UserProfile.ProfileData + your full private-session transcript. Output (ComplaintAssessment.AssessmentJson) is kept for the life of the relationship and not shown to either partner |
| Conflict — proposal (one shared text; a new version after "Not quite") | Both partners' assessments + what each approved under "try next" in their summary + the earlier versions and the one change both agreed to + the topics and agreements of up to three earlier closed Kitchen chats or tables of the same two people, and whether those agreements held + the household context (as in the opening). Output: the shared text with its action items, and a private note to each partner |
| Conflict — the check on a "Not quite" request | The asker's reason and request + both partners' assessments and summaries + the proposal. Output: the request as one neutral sentence for the other person, or a note to the asker alone that holds it back to be rewritten |
| Conflict — impasse summary | The shared proposal texts only |
| Conflict — closing package (a couple's table, once both have signed) | Both partners' assessments. Output: a shared explanation of the pattern the two were caught in and, where one is offered, an exercise for each |
| Kitchen chat → table draft | Only your own messages in the Kitchen chat. Output: a draft of your description, which you edit before anything is sent |
| Quick Repair (brain-gated mediator) | The whole thread + display names + the couple's "About us", or the family block for a parent and a young person |
| Teen check-in distillation (with the teen's consent) | The teen's entries for the week (mood + text per area of life). Output: one theme per area, an opener and a "don't" for parents, and the safety-check verdict |
| Chat turn in a family chat (when @refairly is tagged, or the safety check alone) | Only when someone writes @refairly or a young person asks for help saying something: the recent transcript of the chat, members' display names and roles (parent / young person), the chat kind, and only what every member may see (§4.7). Output: a reply and up to three cards that only the one who asked can tap (no reply when a young person in the chat has not consented) and the safety-check verdict |
| Your Refairly chat turn (an adult's) | Every message: the recent conversation, with the app's record of Refairly's earlier turns there (which reads worked, which messages got no reply); Refairly's notes about you; the trip plans you are on that were made in the last seven days; and what you may see in the app, read as needed (§4.7) — at most three model calls a turn. Output: a reply, up to three cards you tap and, now and then, a note about you kept, changed or forgotten |
| A young person's Refairly chat turn | Every message: the recent conversation, with the app's record of Refairly's earlier turns there (which reads worked, which messages got no reply, and whether the help resources took the place of a reply), and only her own things (§4.7), her recent check-in entries with her words among them. The same call carries the safety check, which decides whether her family is asked to check in, without what she wrote. Output: a reply and, while her teen consent is active and no safety signal was found, up to three cards she taps |
| A young person's goal note | The note's text, for the safety check only (§10.5) |
| Refairly reading a paper, a goal's photo or a file | The one file it needs to answer — an image (scaled down) or a PDF — and only one the person who asks may open; in a chat, only one every member may open |
What Anthropic keeps of these calls, and for how long, is set by its commercial terms (§5.2).
The couple's facts. The facts the couple gave — whether they are married, engaged, dating, living together or co-parents, how they live (together, apart or at a distance), since when they have been together and, while the consent in §2.3 stands, that they are expecting a child, which is health data — go with "About us" in the "About us" synthesis itself, a couple's Kitchen chat and table, the couple's chat when Refairly is asked there, and a trip plan that is only the two of them (the question before a build, the build and the plan's chat); a family's trip plan carries the parents' facts without "About us" and never says that they are expecting. Refairly also reads them in an adult's own Refairly chat when it needs them (§4.7).
4.2 Non-AI external services the Trip Planner calls
The Trip Planner specifically calls additional non-AI providers as part of generating a personalized plan. To build a plan we send these services only the trip details needed for the search: city, area, dates, place types and the stay style you chose. We don't send your messages, names or answers.
| Provider | When | Data sent |
|---|---|---|
| Google Places (New) | Every generation, and when a change asked for in the plan's chat needs another place or another stay | City, country, area lat/lng, search radius, place type categories, and text searches built from them (for example the stay style you chose and the city or area) |
| Google Routes | Every generation + on every plan-mutation tool call | Pairs of (origin lat/lng, destination lat/lng) and the travel mode (for example walking, driving or public transport) |
| Google Place Photo | Every generation that has candidate photo refs | Each photo reference token + our backend API key (server-side only; the mobile app does not call Google) |
| OpenWeather | When the trip starts within the next 8 days | Area lat/lng |
| Ticketmaster Discovery | For a weekend or a short trip, and for a dinner or a day out whose purpose is "celebrate" | City, country, date window |
| Overpass (OpenStreetMap) | Conditional (when the taste of anyone on the plan leans to nature or adventure) | Area lat/lng, radius, OSM tags |
| OpenTripMap | Conditional (when the taste of anyone on the plan leans cultural, or the purpose is "tryNew") | Area lat/lng, radius, POI categories |
Trip Planner photo storage. Photos surfaced from Google Places are fetched server-side, re-encoded to WebP, and cached in our AWS S3 bucket under the prefix trip-photos/. The bucket policy for this prefix grants public-read to anyone with the URL. URLs are content-addressed (a hash of the Google photo reference) and not easily guessable, but they are not access-controlled — anyone who learns a specific URL can fetch the WebP image. Photos are Google's place photos (third-party business imagery), not user-uploaded photos. Each photo is shown with its author's name as Google provides it, in the app and in a shared PDF. We store one photo per place and show it in every plan that offers that place; when it is more than 30 days old we fetch it again from Google. A place offered as a choice gets its photo the first time someone opens that stop's choices.
4.3 Automated decision-making & your rights
The AI generates per-partner behavioral assessments (Conflict layer's ComplaintAssessment), the Kitchen's proposals and its check on a "Not quite" request, plan content, first goals, weekly themes about a young person, Refairly's cards and the notes it keeps about an adult, and the safety-check verdicts described in §10.5. These are algorithmic estimates and proposals, not binding decisions. No fully automated decision with legal or similarly significant effects is made — all parties must voluntarily agree to any proposed outcome; a "Not quite" request the check holds back goes back to the person who wrote it, with a line saying why, to rewrite; a card from Refairly changes nothing until the person who asked taps it (§4.7); a note Refairly keeps can be seen and deleted in Settings → What Refairly remembers (§4.7); and the only automated actions the safety check can take are to show help resources, to pause a young person's Kitchen chat or table and keep her words for her alone, and to send a parent a contentless request to check in.
Under GDPR Art. 22 and KVKK m.11/g, you have the right to:
- Object to outcomes produced solely by automated processing
- Request human review by contacting aoci@refairly.app
- Decline any specific AI proposal (say "Not quite" to a Kitchen proposal, dismiss a card from Refairly, edit or delete a goal Refairly wrote, delete a note Refairly kept about you)
- Report harmful, biased, or incorrect AI outputs for review
There is currently no routine human review of AI outputs. If you report an issue we aim to review the relevant output and respond. A young person can ask for human review of a theme or a safety-check event through their parent or directly at the address above.
4.4 AI safety, silence, and limits
- Silence is a valid output. The Refairly Brain decides whether to speak on Quick Repair and on the Trip Planner's question before a build; on most turns it stays silent. In the chats Refairly speaks only when asked with @refairly; the exception is the chat of a trip plan with no young person with her own account on it, where it answers every message. Silence is not endorsement, agreement, or absence of concern.
- The AI is instructed not to take sides between partners, or between a parent and a young person, and not to disclose one person's private content (Daily Reflection, Conflict private session, individual onboarding, a teen's entries) to another. Instructions guide the model; they are not a guarantee of what it writes.
- No clinical diagnosis or label. The AI is instructed not to produce diagnoses, attach clinical labels, or predict relationship outcomes, and Refairly does not advise whether a medicine dose may be given. Themes about a young person are phrased as needs, not as verdicts.
- Safety monitoring exists only in the family surfaces. The couple features run no server-side keyword detection, abuse-content classification, or self-harm signal extraction; safety-conscious AI behavior there is enforced by AI system prompts only. A young person's check-ins, her own Refairly chat, her goal notes, her Kitchen chat and table, and the messages she sends to Refairly in a chat run the automated safety check in §10.5; others' messages are not checked. If you or someone else is in immediate danger, contact local emergency services.
- Limits on Refairly's cards. Refairly offers only a few cards a turn. A card changes nothing until the person who asked taps it, and a card that removes or ends something asks the same "Are you sure?" its screen asks. In a trip plan's own chat, Refairly changes the plan itself, without a card (§4.7).
- AI non-response. If an AI call fails or the pool is used up, the AI skips that turn; your message is still posted, and the app usually shows a note saying why. In a person's own Refairly chat and in a trip plan's chat, Refairly's later turns are told that the message got no reply, so it can still answer it; that note is for Refairly alone.
4.5 AI disclaimers
AI-generated analyses, plans, suggestions, summaries, themes, and assessments are not legal advice, medical advice, mental-health advice, professional mediation, therapy, or arbitration. Refairly is not a law firm, mediation service, licensed therapy provider, arbitration body, or crisis service. AI output may contain errors, biases, or inappropriate content. Always use your own judgment and consult a qualified professional for matters with legal, financial, medical, or safety significance. See our Terms of Service for full disclaimers.
4.6 Reflection Insights (optional)
If you switch on Reflection Insights in Settings, your private Daily Reflection answers are periodically distilled by Anthropic into a short private digest — an early-warning sense of how you have been doing. The digest is stored on your account (ReflectionDigest). The digest itself is not shown to your partner and is not put into a shared conversation; it is used in your own private AI conversations (your Conflict private session and related private prompts); Refairly, the assistant, does not read it. A rough score drawn from it, together with both partners' relationship-check totals and how your Kitchen check-ins went, sets a coarse trend label (improving or strained) that is given to the AI in the Kitchen chat and the couple's chat to set how readily it speaks; the label carries no words from your reflections. Switching Reflection Insights off deletes the digest, and it no longer counts toward the label. Your reflections themselves are unaffected either way: stored, private, and not sent to the AI without this consent.
4.7 Refairly, the assistant — what it reads and the cards you tap
What it is. Refairly is one assistant across the app. You meet it in your own Refairly chat — every adult account and every young person with her own account has one; caregivers have none — and in a chat when someone writes @refairly (the couple's chat, the family group, a young person's chats with her parents; not the caregiver chats). It answers questions about your family's things and, when you ask for a change, suggests it as a card. In an adult's own Refairly chat it also keeps short notes about that adult, so it knows them next time (below).
Cards, and who taps them. A card shows what would change before anything changes. Nothing changes until the person who asked taps it; in a chat everyone sees the card, and only the one who asked can tap it. The tap does what the app's own screen does, with that person's rights at that moment — the server checks them again, and checks that the thing has not changed since the card was made — and sends the same notifications the screen sends. A card that removes or ends something asks the same "Are you sure?" the screen asks. Refairly does not act on its own — the exceptions are the short notes it keeps about an adult in their own Refairly chat, without a card, which change nothing else in the app (below), and a trip plan's chat, where it changes the plan itself, without a card, from the messages it answers, as far as the writer of the message may: a stop, the order of a day, the plan's answers, a line of things to plan around said there by someone who may write it, and, on an adult's message, the stay or a rebuild (§2.1) — and it does not act for anyone but the person who asked. A card expires if it is not tapped: after 24 hours; a "today" card on the Day at the child's midnight; a recorded dose, a take-back or a message after 10 minutes.
What it can suggest. Almost anything you can do in the app yourself: goals and their text notes; the Day, medicine included, and the Home day; a record that a dose was given or not given when you say it happened (recorded at the moment you tap); alarms on goals and Day items — one for you, or one that asks the people you name (it rings them only once they accept), a change to one or its removal, your own settings on one, and your answer to a request (§2.10); a child's card and health list, and "what you've noticed"; who a paper is for, or removing one; what a caregiver may do, or taking a child back from one; pausing or resuming a young person's account; household expenses, budgets and settling a month; pocket-money entries; your pulse word and who sees it; a message sent as you into one of your chats; connection requests; the steps of ending a family; taste answers; notification settings; your name and app language. A card with an amount, a dose, a time, a gap or a limit is made only from a number you wrote in your recent messages; otherwise Refairly asks you. Some things open their own screen instead of a card: codes and links, files and photos from your phone, calls and buzzes, purchases, passwords and sessions, legal and consent texts, forms of several questions, your check-in, the day's private writing, reports and device settings. The Trip Planner and the Kitchen: with its cards, Refairly does not change them; it explains how to do things there yourself. In a trip plan's own chat it changes the plan as described under Cards above.
A message sent as you. A card can send a message as you into one of your chats (not a caregiver chat or a closed chat). It shows the chat, everyone who will read it and the exact text. It is sent only when you tap, it is sent as you — it is not marked as written with Refairly — and it cannot be unsent.
What it reads. Only what the person who asks may see in the app, and in a chat only what every member of that chat may see. It reads when it needs to answer:
| Where you ask | What Refairly may read | What it does not read |
|---|---|---|
| An adult's own Refairly chat | Your chat with it, with the app's record of its earlier turns there, and the notes it keeps about you; the household's goals and the plan's reading; your own goal items and your partner's; your children's goals with their notes; the Day as you see it and the household's Home day; the alarms on the items and goals you see, the alarm requests waiting for you and your next alarms; your children's cards, including the health list (health data), and both parents' observations; the themes a young person chose to send and her mood numbers; her pocket money; this month's expenses and budgets; "About us" and the last agreements you both signed (their title and what was agreed); the pulse words you may see; the chats you are in, except the caregiver chats — one with a young person only while every young person in it has her teen consent active — and your notifications inbox; a paper, a goal's photo or a file you may open, when it needs one; your taste answers; the names and status of your trip plans and of your Kitchen chats and tables (not what is written in them), and, with every message, the trip plans you are on that were made in the last seven days — what kind of plan, where, the dates, who is on it, how far along it is and when it was made; the relationship's facts, its timeline and moments, and connection requests; where a family ending stands; your account, your consents, and your plan and its pool's use; your own day, your medicines only while your own-health consent stands; your partner's day, which you see in full (their medicines only while their own-health consent stands); a mother's or father's card, health list, day, medicines, goal items, notes and papers once they have agreed in person to Refairly reading their record; a pet's card and day | Anyone else's Refairly chat; a young person's check-in words, even those she shares with you; your daily reflections and their Reflection Insights digest; a Kitchen private session, its summary or its messages; a trip plan's chat; satisfaction-check scores; chats you are not in, and the caregiver chats |
| A young person's own Refairly chat | Her chat with it, with the app's record of its earlier turns there; her goals with their notes; the Day items she sees; the alarms on them, the alarm requests waiting for her and her next alarms; her pocket money; her recent check-in entries, with her words and her mood numbers; the themes she sent and whether she checked in this week; the chats she is in (not her chat with her caregiver) and her notifications; the pulse words she may see; her taste answers; her account and consents; the papers meant for her, when it needs one | The household's plan, goals, Home day, money and agreements; her parents' own goal items; her parents' observations and her card; her brothers' and sisters' records; papers kept to her parents, including their names |
| The couple's chat | What both partners may see: the household's goals and its Home day, each partner's own goal items, the goals of children you both parent and their cards, health list included, with your observations about them; this month's money; papers both may open; the themes and mood numbers of a young person you both parent; "About us" and the last agreements you both signed; the alarms on goals and Day items both may see; each partner's own day (medicines only while that partner's own-health consent stands); a mother or father once they have agreed to Refairly reading their record; the pet | Either partner's own Refairly chat, and the notes Refairly keeps about either of them |
| A chat of parents with a young person | What everyone in it may see: her goals, her Day as she sees it (never an item her parents keep to themselves) and the alarms on them, her pocket money, her themes and mood numbers, papers she may open | The household's data and money — its goals and Home day included — her parents' own goal items, her brothers and sisters, her card and health list, her parents' observations |
| The family group | Only what every member may see: nothing of the household when a child is in it; a child's goals only when no brother or sister is in it | Anything one member may not see |
| Caregiver chats | No Refairly | — |
A child's things and the children's-data consent. A parent's Refairly reads anything about a child — the record and its card, the health list, observations, the Day and its medicine records, goals with their notes and files, papers, pocket money, caregivers, the young person's account, the themes and numbers she sends and her pulse word, and notifications about the child — only while that parent has given the children's-data consent (§2.3) to its current text; in a chat, it is the consent of the person who asks. Without it, Refairly reads nothing new about that child for them, makes no card about the child, and points them to the child's page, where the consent is given; in a chat the young person is in, it says only that their own Refairly chat explains. Replies it gave before stay in the chat's recent history. A young person's own Refairly works under her teen consent (§2.3).
Health data. When a parent of the child asks — in their own Refairly chat, or in a chat whose every member is a parent of that child — Refairly may read the child's health list and the Day's medicine records, which are health data, and a health report among the child's papers when it needs one: only while that parent has given the children's-data consent and, in a chat, the health list and a health report only while every parent in it has. It reads them to answer that person or to make a card that person taps; in a chat, everyone in it sees the reply. It does not advise whether a dose may be given.
Your own medicines, and Mom or Dad. Refairly reads an adult's medicines only while that adult's own-health consent stands, for that adult or their partner. It reads nothing of a mother's or father's record until their son or daughter has recorded that they agreed in person to Refairly reading it; after that, only for people who may see them (§10.15). What someone writes about them in a chat is sent as that person's words. When a consent or an answer is withdrawn, Refairly reads nothing new; its earlier replies stay in the chat's recent history, which goes to Anthropic with each new message. In a chat with a young person or a caregiver in it, it reads nothing of a partner's own day. Caregiver chats, a chat about Mom or Dad included, have no Refairly. Refairly can add a pet by a card the person who asked taps; it never adds a mother or father, and never records or takes back their answers — for those it opens the app's own form or their page.
What Refairly remembers. In an adult's own Refairly chat, in app versions that show these notes (Settings → What Refairly remembers), Refairly keeps short notes about the person it is talking with, so that it knows them next time. It keeps one by itself, without being asked and even from something said in passing, when they tell it something lasting about themselves — how they like to be called, their work and its hours, their routine, what they like or avoid, how they like its answers — and when they ask it to remember something. It writes a note only from what that person wrote in that chat — never from what it read in the app or in a file — as one sentence of at most 300 characters, and keeps at most 30 about one person; with 30 kept, a new note replaces one. It keeps, changes and forgets notes by itself, without a card; when a reply kept or changed one, the app says "Noted" under it, with a way to the notes. It is instructed never to keep a detail about health, religion or beliefs, sex life or sexual orientation, ethnic origin, political views or union membership: one about the person themselves only when they ask it to remember it — that request is their explicit consent to keeping it (§3) — and one about anyone else never. It is also instructed never to keep anything from the Kitchen or about a conflict with someone, another person's private matters, or a password, a code, or an ID, card, account or phone number. These are instructions to the model, not a filter: a note can be wrong, or keep more than it should, so do not ask Refairly to remember another person's health or private matters. The person sees every note in Settings → What Refairly remembers and can delete any of them, or all; a deleted note is gone for good, and clearing the chat does not delete them. The notes are read into each turn of that person's own Refairly chat, so they go to Anthropic with each message there; they are read nowhere else — not in a chat with other people, not in a partner's Refairly chat — and change nothing else in the app. Refairly keeps no notes about a young person or a caregiver. While the AI permission is withdrawn, Refairly does not answer, so it neither reads nor keeps notes; the ones it kept stay until the person deletes them or their account. They are in the data export (§9.1).
Papers, photos and files. Refairly reads one when it needs it to answer: an image (scaled down) or a PDF; it does not read Word or HEIC files. It reads only a file the person who asks may open; in a chat, only one every member of the chat may open. A young person's Refairly does not see papers kept to her parents, or their names. The file is sent to Anthropic for that answer.
A young person's Refairly. Her cards need her teen consent to be active, and no card is made on a turn where the safety check finds a signal or the pool is empty. The safety check runs on what she writes (§10.5).
Where it runs and who pays. Claude, made by Anthropic, as our processor (§5.2). A turn uses at most three model calls. It is paid from the household's pool — for a young person, a parent's — and recorded as Refairly's use in the usage log, without content.
What we keep. Each card is stored as a RefairlyAction: who asked, in which chat, what it would change, its state (suggested, done or dismissed) and when. When a child's record is removed — or the record of Mom or Dad or a pet — each card about them — their card and health list, their papers, their caregivers, their goals and their notes, their Day and the alarms on it, a child's pocket money — stays under Refairly's reply as an empty "no longer available" placeholder, and so does a card that quoted a young person's goal note when her account is deleted, and a card about a deleted goal or a removed Day item when it is erased, 30 days after its removal. A placeholder keeps only what kind of card it was, whose it was, where it sat and when; nothing it quoted is kept, and nothing can be done with it. It goes when the reply it sits under goes. Each note Refairly keeps about an adult is a RefairlyNote: its text, and when it was kept or last changed. On each message in a person's own Refairly chat we keep the app's record of that turn: each read Refairly ran, whether it worked and, when it did not, a fixed reason or the app's error code, and whether the help resources took the place of a reply — never what a read returned, and nobody's words. It is read back into that chat's later turns, for Refairly alone, is shown to nobody and goes with the message; whether the help resources took the place of a reply is removed from it 365 days later, with the safety-check records (§10.5). A reply also records whether its turn kept a note. Retention is in §8.
5. Sub-Processors & Third-Party Services
Refairly uses the following third parties to deliver the Service. A change in provider, model, SDK, or underlying technical implementation does not by itself change the purposes for which we process your data; we update this list as providers change.
5.1 Hosting, storage & monitoring
| Service | Provider | Location | Purpose | Data involved |
|---|---|---|---|---|
| API host | AWS Lightsail | EU (Frankfurt) | Compute, single-instance .NET API + nginx TLS | All inbound API traffic |
| Website | AWS S3 and CloudFront | Files stored in the EU (Frankfurt); pages served from AWS's edge locations around the world | Serving refairly.app, including these documents and the account-deletion page | Your IP address and the address of the page you open, to answer the request; we keep no access logs for the website |
| Primary database | AWS RDS for PostgreSQL | EU (Frankfurt) | All structured application data; encrypted at rest via AWS-managed encryption | All entities listed in §2.1 |
| File storage | AWS S3 | EU (Frankfurt) | Receipt photos, a child's papers and goals' photos and files (private, encrypted at rest, served through short-lived signed links); profile pictures (private, encrypted at rest, served through signed links that expire after a day); cached Trip Planner photos | Receipts, papers, goal files and profile pictures in three JPEG sizes (private); Trip-Planner cached place photos (public-read under trip-photos/) |
| Server error monitoring | Sentry (Functional Software, Inc.) | EU (Germany) data region | Capturing server errors so we can fix them | Error type and stack trace, request path and method, a user identifier where our code attaches one, technical log lines. Designed not to include message bodies or AI prompts; email addresses are masked |
5.2 AI providers
| Service | Provider | Location | Purpose |
|---|---|---|---|
| All AI inference (Conflict / Quick Repair / Growth / Trip Planner / Joint onboarding / Reflection Insights / teen check-in distillation / family chats / the family safety check / Refairly, the assistant — its chats and the notes it keeps about an adult, its tagged turns in the couple's and family chats, its cards, and reading papers, goal photos and files when it needs one) | Anthropic (Claude models; which model runs which feature may change) | USA or other countries | LLM inference per §4.1. Anthropic processes this data as our service provider under its commercial terms. Under those terms it does not use it to train its models, and it keeps inputs and outputs only for a limited period for its own safety and legal purposes. So that it can answer faster and at lower cost, the parts of a request that repeat from one call to the next (the instructions, and the context a chat sends with each message) are also held by Anthropic for a few minutes after each use, so the next call can reuse them; they are kept apart from other customers' data and used for nothing else. See Anthropic's privacy terms |
We do not currently route any AI inference through Google Gemini, OpenAI, or any other LLM provider.
5.3 External APIs the Trip Planner calls (see §4.2 for details)
| Service | Provider | Location |
|---|---|---|
| Places search | Google Places (New) | United States / global |
| Routing | Google Routes | United States / global |
| Place photos | Google Place Photo API | United States / global |
| Weather forecast | OpenWeather | United States / global |
| Events | Ticketmaster Discovery | United States / global |
| Public-domain map features | Overpass (OpenStreetMap) | EU / community-hosted |
| Tourist points-of-interest | OpenTripMap | EU |
5.4 Communication, push, calls & analytics
| Service | Provider | Location | Purpose |
|---|---|---|---|
| Transactional email | Resend | United States | Account emails (welcome email, email confirmation, password reset, account-deletion verification code), and internal alerts to us about safety reports and "Report a problem" messages, which carry what was reported and, for a safety report, the reporter's comment and IP address |
| Support mailbox | GoDaddy (email hosting for refairly.app) | Not stated by GoDaddy | Holds what you write to aoci@refairly.app, and the copies of safety reports and "Report a problem" messages emailed to us (§2.1, §8) |
| Second mailbox for safety reports | Microsoft (Outlook.com) | United States / global | A copy of each safety report is also emailed to a mailbox of ours at Outlook.com (§2.1, §8) |
| Push (Android + most iOS) | Firebase Cloud Messaging | United States | Push delivery, including silent data pushes (among them the one that tells a phone its alarms changed, which carries no content — §2.10) |
| Push (iOS Buzz Call invitations only) | Apple APNs (direct connection) | United States | High-priority VoIP push for incoming buzz calls |
| Call connection setup (voice in Buzz Calls) | Google STUN servers | United States / global | During a Buzz Call your phone contacts Google's public STUN servers to learn its own public address, so the call's audio can go directly between the two phones (§2.11); those servers see your IP address. No call audio passes through them. We use no relay (TURN) server today; one added later will be listed here before it is used |
| Analytics | Google Firebase Analytics | United States | Usage events, screen views; may include user identifiers. Off until you turn it on — when you sign up, on the documents screen or in Settings — and you can turn it off in Settings at any time; never on for a teen account. When you give a consent, it records only that a consent was given, not which one. On the first launch after updating from an earlier version on which analytics was on, Firebase's own start-up events (such as the start of a session or the app update) may still be sent once, before the app turns it off |
| Crash reporting | Google Firebase Crashlytics | United States | Crash logs, device info; may include user identifiers. Error messages sent are limited to technical descriptions |
| App updates | Expo (650 Industries, Inc.), EAS Update | United States | Each time the app starts, it asks Expo's update service whether a newer version of the app's own code is waiting, and downloads it if so. Expo sees the phone's IP address, its platform, the app's version and a random identifier of this installation; it receives no account details and nothing you wrote |
| Social sign-in | Google OAuth / Apple Sign-In | United States | Authentication (adult accounts) |
| In-app purchases — iOS | Apple App Store + Apple App Store Server API + Apple App Store Server Notifications V2 webhook | United States / global | Payment, receipt verification, renewal/refund webhooks |
| In-app purchases — Android | Google Play Store + Google Play Developer API + Google Cloud Pub/Sub RTDN | United States / global | Payment, receipt verification, settle via :acknowledge / :consume, renewal/refund webhooks |
6. Cross-Border Data Transfers
Our primary infrastructure is in the European Union (Germany / Frankfurt), including our error monitoring. Certain data is transferred to the United States, and for some providers to other countries where they run their services (§5), for processing by the providers listed in §5 — among them Anthropic, our AI provider, which may process what we send it in the USA or other countries where it runs its service; Microsoft, whose Outlook.com mailbox receives a copy of each safety report; and GoDaddy, which hosts our support mailbox.
6.1 EU/EEA users
Transfers to sub-processors in the United States, and to the other countries where some of them run their services (§5), rely on one or more of: (a) Standard Contractual Clauses (SCCs) adopted by the European Commission; (b) the EU-U.S. Data Privacy Framework where the recipient is certified; (c) the provider's own adequacy mechanism as documented in their data-processing agreements.
6.2 Turkey (KVKK)
For users whose personal data is subject to KVKK, storing data in the European Union and transferring it to service providers outside Turkey are transfers abroad under KVKK m.9. Our regular transfers are to Anthropic, Amazon Web Services, Google, Apple, Resend, Sentry, Expo, GoDaddy and Microsoft. For regular transfers without an adequacy decision, KVKK m.9 asks for an appropriate safeguard, such as the standard contract the Kurul has published; when we sign standard contracts with these providers, we will state their type and date in this section and in the KVKK notice. If Turkish law or Kurul guidance changes, we may update our transfer mechanisms and this section accordingly.
6.3 UK users
For users in the United Kingdom, transfers are governed by the UK GDPR and the International Data Transfer Agreement (IDTA) or UK Addendum to EU SCCs, as applicable.
6.4 Canadian and Quebec users
Your personal information is stored in Germany and processed by the providers in §5, some in the United States or other countries. Before transferring personal information outside Quebec we assess, as Law 25 requires, the sensitivity of the information, the purposes, the protections in place and the legal framework of the destination. You may ask us about it.
7. Data Security
We use reasonable technical and organizational measures designed to protect your personal data. No method of transmission or storage is completely secure, and we cannot guarantee absolute security.
Our measures include:
- Encryption in transit and at rest, hashed passwords and sign-in tokens, access controls checked on our servers, limited staff access, rate-limiting, and automated retention and deletion (§8)
- The visibility wall between a young person and their parents is enforced on the server, not in the app: what a parent's request returns of a young person's check-ins follows §10.3, and a request from a teen account does not reach a couple feature
- Refairly's cards are checked again on the server when tapped: only the person who asked can apply one, with their rights at that moment
- Who an alarm rings is decided on the server: only people who can see what it is on, and nobody on someone else's request until they accept it (§2.10)
Our security measures evolve as systems, vendors, and industry practices change.
8. Data Retention
The following table is the canonical retention reference. Periods are applied by an automatic daily process, so data is deleted on or shortly after the date shown.
| Data type | Retention period | Behavior on relationship end | Behavior on account deactivation (deletion requested) | Behavior on permanent deletion (≥30 days after the request) |
|---|---|---|---|---|
Account row (AspNetUsers) |
While account is active | — | Display name removed, sign-in tokens revoked, IsActive=false, security stamp rotated; email kept only so the account can be restored within the 30 days |
Hard-deleted |
UserProfile |
While account is active | — | Preserved | Hard-deleted |
UserMemory |
While account is active | — | (no explicit soft-delete pass) | Hard-deleted |
| Consent records (every scope, and legal-document acceptance: what you agreed to, the text version, when, and the IP address and device used) | While account is active, and after it | — | Kept | Kept after the account is deleted, to prove consent was given, for as long as we may need that proof under the law |
Sign-in sessions (UserRefreshTokens) |
Until logout, expiry, or revocation; revoked and expired rows are pruned after 30 days | — | All revoked immediately | Hard-deleted |
Relationship |
While both partners exist | Status flipped to Ended; row preserved |
(no row mutation) | Row hard-deleted, cascade clears all per-relationship child tables (Conflict, Quick Repair, Growth, Daily, Trip, Joint Onboarding, Buzz, Buzz Calls, words for the day, relationship facts) |
ConnectionRequest |
Indefinitely | — | (no explicit soft-delete pass) | Hard-deleted |
JointOnboardingResponse + Relationship.SharedContextJson |
While relationship exists | Preserved on Ended status | Preserved | Cascade-deleted with Relationship |
DailyCheckIn (private answers) |
While account is active | Preserved on Ended status | Preserved | Cascade-deleted with Relationship; one written in no relationship (by an adult with no partner who looks after someone) is hard-deleted |
ReflectionDigest (opt-in) |
Until you withdraw Reflection Insights | Preserved | Preserved | Hard-deleted |
PartnerPulse, PulseAudience |
While account is active | Preserved on Ended status | Preserved | Deleted with your account |
The couple's chat (DailyChat of kind couple, DailyChatMessage) |
While the relationship exists | Preserved on Ended status: the chat closes, stays readable and takes nothing new | Your messages leave the chat at once, as you do (removed, not yet erased); the rest of the chat stays for the other members | Cascade-deleted with Relationship |
Growth Plan (GrowthPlan, GrowthPlanResponse, the plan's reading) |
While the household exists: a couple's relationship, or a parent's own account | Preserved on Ended status. With children, it moves to the parent who keeps the family. | Preserved | Cascade-deleted with Relationship (or with its owner's account) |
Household goals (Goal on the household's list) and their notes, photos and files (GoalNote, files on S3) |
While the household's plan exists. Removing a note deletes it for good with its stored file; a deleted goal leaves the list at once with its notes and files, and is erased 30 days later | Preserved; with children, they move with the plan to the parent who keeps the family | Preserved | Deleted with the plan: with the relationship, or with a solo parent's account — the rows and the stored files |
An adult's own goal items (Goal on an adult's own list) and their notes, photos and files |
While the adult's account exists. Removing a note deletes it for good with its stored file; a deleted goal leaves the list at once with its notes and files, and is erased 30 days later | They stay with their owner; the former partner no longer sees them | Preserved | Deleted with the owner's account — the rows and the stored files |
A child's goals (Goal on a child's list) and their notes, photos and files |
While the child's record exists. When the child's record is deleted, the goals' notes and stored files go at once; the goals themselves are kept out of everyone's reach and erased with the record, 30 days after it was deleted. Removing a note deletes it for good with its stored file; a deleted goal leaves the list at once with its notes and files, and is erased 30 days later | Unaffected (they belong to the child's record) | Preserved | A young person's own notes and files are deleted with her account; goals and notes an adult wrote stay on the child's list, shown without their name |
The former plan chat (GrowthPlanChatMessage) and skills (ChildSkill) |
Removed in the family release: the plan chat and every conversation it held, and the skills, were deleted. Steps became goals | — | — | — |
Refairly's cards (RefairlyAction) |
Kept with the chat the card was made in. A card that was dismissed or expired is deleted after 30 days; a card that was tapped stays as part of that chat. Clearing your Refairly chat dismisses its open cards. A card about something removed with a child's record is emptied at once, and one about a deleted goal or a removed Day item when that is erased, 30 days after its removal — it stays as a "no longer available" placeholder holding nothing it quoted — and is deleted once the reply it sits under is gone (§4.7) | Unaffected (a card belongs to the person who asked) | Kept, so a restored account gets them back | Deleted with the account of the person who asked |
Trip Planner (TripPlan, TripPlanParticipant, TripPlanContent, TripPlanChatMessage, TripPlanItemInsight) |
While the household exists (a couple's relationship; a solo parent's account); closed plans stay readable. A line of things to plan around is deleted when the plan ends, or 7 days after the trip's last day | Preserved on Ended status | Preserved | Deleted with the relationship or the solo parent's account; a participant's row and line go with their own account |
Place photos (TripPlacePhoto: a place id, one photo and its author's credit from Google; no personal data) |
Shared by every plan that shows the place; fetched again from Google when more than 30 days old | Unaffected | Unaffected | Not tied to any account or household |
Trip Planner S3 photo cache (trip-photos/) |
Content-addressed; one object per unique Google photo reference, possibly shared across many plans | Not tied to a single relationship — not cleared on relationship end | Not cleared | Not currently pruned even on hard-delete; objects are public-read business photos from Google |
ExperienceTasteProfile |
While its owner's account exists (a child's: while the child's record exists) | Unaffected | Preserved | Deleted with the account or the child's record |
Relationship check (CoupleSatisfactionSample, MeasurementWindow) |
While the relationship exists | Preserved on Ended status | Preserved | Cascade-deleted with Relationship, and with your account |
All Conflict entities (ConflictSession, ConflictParticipant, ConflictDescription, PrivateSessionMessage, ComplaintAssessment, ConflictSharedSummary, ConflictPrivateSuggestion, ResolutionProposal with each person's answer, ResolutionAgreement with its signatures, ConflictTask, ConflictCheckIn with each person's answer, ConflictClosingPackage, AgreementReviewSchedule and the rest of a table's rows), and the Kitchen's memory of closed chats and tables (ConflictMemoryEntry, with each person's private layer, ConflictMemoryPrivateLayer; §4.1) |
A couple's: while the relationship exists. A parent's and a young person's: until either of them deletes their account | Preserved on Ended status — including private session messages, ComplaintAssessments, and any signed agreements. A parent who leaves the family: their tables with a young person close and stay readable to both | Preserved | A couple's: cascade-deleted with Relationship; each person's private layer of the memory also goes with their own account. A parent's and a young person's: deleted with either one's account |
Quick Repair (QuickRepair, QuickRepairMessage) |
A couple's: while the relationship exists. A parent's and a young person's: until either of them deletes their account. Closing a thread does not delete content | Preserved on Ended status. A parent who leaves the family: their Kitchen chats with a young person close and stay readable to both | Preserved | A couple's: cascade-deleted with Relationship. A parent's and a young person's: deleted with the parent's account, and with the young person's once she has written in it |
Buzz (envelope + status + timestamps) |
Deleted after 365 days, or earlier with the account | Preserved on Ended status | The buzzes you sent are soft-deleted immediately | Deleted with your account (the buzzes you sent), and with the relationship |
BuzzCall (metadata only: who called whom, in which chat, whether it was a voice call, times, outcome and duration — intensity samples and call audio are not stored, §2.11) |
A finished call's record is deleted 365 days after the call started, or earlier with the account. A voice call's set-up messages are held in server memory only while the call lasts | Preserved on Ended status | Soft-deleted immediately, for every call you were on | Deleted with your account (every call you were on), and with the relationship |
Child record (Person, HealthEntry, Keeper, parents' observations in KeeperNote) |
While at least one parent keeps it. Either parent can delete the record when no teen account is attached to it, which deletes it for both at once: its stored files go then, and the rest is kept out of everyone's reach and erased 30 days later (the expenses that were only for the child stay hidden in the household's money and are deleted with it) | With children, one parent carries the family on and the parent who leaves loses access (§8.1) | Preserved for the other parent, and your parent link is removed; if you were its only parent, the record is deleted then — unless a young person has her own account on it, which stays hers — and restoring the account does not bring it back | Erased |
| Child-scoped expenses, receipts, budgets, pocket money (a parent's top-up and the household expense that mirrors it are removed together) | With the child record / the relationship; when the child's record is deleted, the expenses that were only for the child are kept out of everyone's reach and deleted with the household's money | As above | As above | Receipts removed from storage with the expenses that go with your account (a receipt on an expense that stays with the family stays with it); rows deleted with the record or relationship |
The Day (DayItem and its changes, DayOutcome, reminders) |
While the child's record exists. An item removed from it is kept out of sight and erased, with its changes, records and reminders, 30 days after its removal | As the child record | As the child record | Deleted with the child's record |
The Home day (its Person row, its DayItem and DayOutcome rows) |
While the household exists. An item removed from it is erased 30 days after its removal | With children, it goes with the family to the parent who keeps it; without children, neither former partner sees it any more, and it stays in each one's data export (§9.1) | Preserved | Deleted with the relationship, or with a solo parent's account, with the alarms on it |
Your own day (DayItem, DayOutcome on your own person row) |
While your account exists. A medicine deleted from it is deleted for good at once, with its records; withdrawing your own-health consent by writing to us, or with the switch of earlier app versions, deletes every medicine and their records at once (deleting your last medicine yourself withdraws it, and also erases for good the medicines removed from your day earlier and kept out of sight). A medicine changed into another kind of item (a walk, a meal) keeps its past as a medicine — its earlier names and notes and its records of taken or not taken — until your consent is withdrawn, in any of these ways, which erases that past; deleted by you, such an item goes for good with its past. Any other item deleted from your day — such an item too, when your partner deletes it — is kept out of sight and erased 30 days later | Stays yours; your former partner's access ends at once | Kept, so a restored account gets it back | Deleted with your account |
Mom or Dad's record (their Person and everything under it: health list, day, goal items, notes and papers, caregiver links) |
While their son or daughter keeps it; they can delete it at any time, and a deleted record is erased from our systems 30 days later, as is an item removed from their day or a goal item deleted. Their health list, medicines and papers are deleted at once when they change their mind, and so is the past as a medicine of one of their medicines changed into another kind (§10.15) | Stays with their own son or daughter, even when that is the one who leaves; the other partner's access ends at once; their caregiver chat stays with their son or daughter | Deleted when their son or daughter asks to delete their account; restoring the account does not bring it back | Erased |
| The record of their answers (who recorded each, when, the words' version; no health data) | Kept like other consent records, to prove the agreement, for as long as we may need that proof under the law, including after their record is deleted | — | Kept | Kept |
A pet's record (its Person and everything under it) |
While a keeper keeps it; either keeper can delete it, and a deleted record is erased from our systems 30 days later, as is an item removed from its day | Goes with whoever keeps the family; a couple with only a pet chooses who takes it | If you were its only keeper, deleted then, and restoring the account does not bring it back; otherwise your keeper link is removed | Erased if you were its only keeper; otherwise it stays with the other keeper |
| Expenses for Mom or Dad or a pet | With the household's money. Deleting Mom or Dad's record takes their name off the expense, not the expense. Deleting a pet's record takes the expenses only for the pet in your own household's money, as a child's does; any other line stays, without its name | With whoever keeps the family; the person's name on those lines shows only to the people who can see them | As household money | As household money |
Alarms (Alarm, AlarmPerson; §2.10) |
While what it is on exists and someone is on it. A one-time alarm is deleted 30 days after its date. An alarm goes with the Day item, goal or child's record it is on, and when nobody is left on it; a person who can no longer see what it is on comes off it. An alarm on a goal that is done, or on a Day item that has ended, rings nothing and stays until it or what it is on is removed | A partner who no longer sees what an alarm is on comes off it; without children, the household goals' alarms go; with children, they stay with the parent who keeps the family | Kept; an alarm rings you only while you can see what it is on. A request you sent that nobody answered is withdrawn. Your devices are signed out, and an alarm already scheduled on a phone can still ring there until the app is next opened on it | Your place on every alarm is deleted; an alarm you set stays for the people on it, without your name, and one left ringing nobody is deleted. If you were a child's only parent, the alarms on the child's things go with the record |
Teen entries (ChildWellbeing) |
While the teen account exists. Her parents see what §10.3 describes; pausing the account does not open more of them to a parent | — | The teen's own deletion removes them | Hard-deleted |
Themes (ChildTheme) |
While the child record exists; deleted when the young person withdraws the teen-wellbeing consent and when the teen deletes their account | — | Removed when the teen deletes their account | Deleted with the record |
Safety-check records (TeenCrisisEvent) |
The AI's description is blanked after 90 days; the record is deleted after 365 days, and so is the mark on her Refairly chat's turn record that help resources took the place of a reply (§4.7); deleted earlier when the young person deletes their account | — | — | Deleted |
ClaimCode |
A code: 15 minutes to redeem; a link: 24 hours. Used and expired codes and links are pruned after 30 days | — | — | Deleted |
Caregiver links (Caregiver; §10.13) |
While the child's record and the caregiver's account exist. When a parent takes a child back, or the caregiver leaves the family, the link is kept as history, marked as removed; a caregiver taken back is offered to the parents again as "looked after before" (one who left herself is not) | Unaffected (the links belong to the child's record) | Kept | Deleted with the child's record, the history too, and with the caregiver's account. When a parent's account is deleted, their id is removed from the links that stay with the other parent |
| A teen or caregiver account whose email was never confirmed and that holds no child | Deleted after 31 days | — | — | — |
Family chats and messages (DailyChat, DailyChatMessage) — the family group, a young person's chats with her parents, caregiver chats, a young person's chat with her caregiver |
While the chat exists; a family's chats go with the child record or the family they belong to; a caregiver chat closes with the last child taken back | The family's chats stay with the family; a parent who leaves the family leaves its chats | Your messages leave the chat at once, as you do (removed, not yet erased); the rest of the chat stays for the other members | Deleted with the record or relationship |
Your Refairly chat, an adult's or a young person's (PrivateChatMessage) |
Until you clear it — "Clear this chat" deletes every message at once, with the record of each turn kept on it (§4.7), and dismisses its open cards; it does not delete Refairly's notes about you (next row) | Unaffected; the chat is yours, not the relationship's | Kept, so a restored account gets it back | Deleted with your account |
Refairly's notes about you (RefairlyNote; adults only) |
Until you delete one, or all, in Settings → What Refairly remembers, or Refairly forgets or replaces it; clearing your Refairly chat does not delete them | Unaffected; the notes are yours, not the relationship's | Kept, so a restored account gets them back | Deleted with your account |
Safety reports (SafetyReport) |
Reporter IP address and user-agent dropped after 90 days; the free-text comment after 365 days; the report record itself is kept as a record of what was reviewed. The emailed copies in our two mailboxes (§2.1, §5.4) are not covered by this automatic deletion; we delete them when they are no longer needed | Preserved | Preserved | Preserved without the reporter's identity link |
| Report a problem messages | Deleted with your account. The emailed copy in our support mailbox is deleted when it is no longer needed | — | Preserved | Deleted |
UserSubscription |
While account is active | Unchanged: each partner keeps the plan and bonus tokens their own account holds | Soft-deleted, so a restored account gets it back | Hard-deleted |
AppAccountTokenMap (store pseudonym) |
While account is active | Preserved | Preserved | Hard-deleted |
TokenUsageLog (no prompt content — counts/cost/provider/model/request-type) |
400 days, or earlier with the account | Preserved | Soft-deleted | Hard-deleted |
| Store notification records (Apple / Google) | Kept for billing, tax and fraud-prevention purposes, for as long as we may need them under the law, including after account deletion; the raw notification payload of a successfully processed event is removed after 90 days | Preserved | Preserved | Not deleted (rows are keyed on store transaction IDs, not user IDs) |
| Plan change history | Kept for billing, tax and fraud-prevention purposes, for as long as we may need them under the law, including after account deletion | Preserved | Preserved | Not deleted — it then points to an account that no longer exists |
NotificationHistoryEntry (in-app inbox) |
Read entries deleted after 90 days, all entries after 180 days, or earlier with the account. A caregiver's entries about a child she no longer looks after — taken back by a parent, a family she left, or a record deleted — and about a chat of hers that closed leave her inbox, its count and her data export at once | — | Soft-deleted | Hard-deleted |
NotificationPreferences |
While account is active | — | Soft-deleted | Hard-deleted |
DeviceToken |
Until revoked / unregistered / terminal push failure | — | Soft-deleted | Hard-deleted |
| Apple / Google OAuth provider linkages | While account is active | — | Cleared on hard-delete | Hard-deleted |
| Profile pictures and receipt photos on S3 | While account is active; when you change or remove your profile picture, the earlier one is deleted from storage | — | Preserved | We delete them from storage: your profile picture in every size, and the receipts on the expenses that go with your account (a receipt on an expense that stays with the family stays with it). A picture an earlier deletion left in storage is deleted when our daily check finds it |
| Deactivated account row | ~30 days from account-deletion request | — | Restorable within the window (§9.2) | Deleted by an automatic daily process |
| In-process rate-limit ledgers (Buzz, Buzz Calls, password reset, etc.) | In-memory only; not persisted; cleared on app restart | — | — | — |
| Sentry error events | Per Sentry's retention for our plan (currently 90 days) | — | — | Not linked back; may still reference your former user id |
| Server logs and IP addresses | Our API's own logs and the server's system journal: deleted after 30 days, and the server's daily snapshots, which hold a copy, are kept 7 days more; the web server's access logs: rotated on a rolling basis | — | — | — |
| Billing / transaction records | As required by applicable Turkish tax and legal compliance obligations | — | — | — |
| Legal hold / fraud investigation | As required by law or investigation | — | — | — |
8.1 Per-user purge inside a multi-party context — what is and isn't supported
We do not currently implement a "delete my contribution from this Conflict / Trip Plan / Growth Plan / family chat while keeping the others' contribution intact" mechanism. Goals are the exception: anyone who keeps a list can delete a note, a photo or a file on it for good, with its stored file, and can delete a goal, which leaves the list for everyone at once with its notes and files; the goal's own title, description and progress are kept out of everyone's reach and erased 30 days later. An item removed from a day — a child's, your own, the Home day, Mom or Dad's, a pet's — is kept out of sight the same way and erased, with its records, 30 days after its removal; a medicine on your own day goes for good at once (§2.12). The closest controls you have today for the rest are:
- End your relationship — flips the Relationship to
Endedstatus. Most per-relationship data is preserved at this point (Conflict descriptions, private session messages, Conflict shared summaries that were approved, Quick Repair threads, Growth Plan content, Trip Plan content, Joint Onboarding answers all remain stored, and either former partner can still retrieve them from their own account — ending the relationship removes them from the app's screens but does not lock them — until either partner closes their account). Each partner keeps the subscription their own account holds (the tier belongs to the account); your word for the day no longer reaches your former partner. With children or a pet both of you keep, the relationship can't simply end: one of you carries the family on. Asking to take the family needs the other partner's acceptance; leaving the family to the other does not need anyone's, so either of you can still end it at any time. The partner who keeps the family continues as a solo parent with the children, their accounts and their chats; the other stops being their keeper in the app, leaves those chats and loses access to their records. A couple with only a pet chooses who takes it. A mother or father one of you looks after stays with their own son or daughter, and the other partner's access ends at once. Each of you keeps your own day. What the leaving partner wrote (observations, plan answers, goals and notes, expenses) stays with the family and is not deleted; their own entries stay in their data export. Refairly does not decide custody. With children, the household's plan, its goals, ledger and moments move to the parent who keeps the family. - Delete a child record — either parent can, as long as no teen account is attached to it; it removes the record with its card and health list, observations, the child's goals with their notes, photos and stored files, the Day and its medicine records, the papers and their stored files, the expenses that were only for them (one shared with a sibling stays, without their name), pocket money and the household expenses that mirror it, what the parents answered about what the child enjoys, themes, safety-check records, claim codes, the alarms on the child's items and goals, every caregiver's link to the child — including the record of who looked after the child before — and the child's chats, for both parents at once; the stored files go then, and the rest is kept out of everyone's reach and erased from our systems 30 days later (the expenses that were only for the child stay hidden in the household's money and are deleted with it); Refairly's cards about the child are emptied and stay only as "no longer available" placeholders (§4.7). While a teen account exists the record cannot be deleted: a parent can pause the account, and once the young person deletes their own account the record returns to a plain record.
- Delete Mom or Dad's record — their son or daughter can, at any time. It removes their card and health list, their day and medicines with their records, their goal items, notes and papers with their stored files, their caregivers' links and their caregiver chat, and the moment that said they joined the family, and the record is erased from our systems 30 days later; the record of their answers is kept, without health data (§8). Expenses for them stay in the household's money, without their name.
- Delete a pet's record — either keeper can. It removes the pet's card, day, notes and papers with their stored files, its caregivers' links, and the expenses only for the pet in your own household's money (one shared with someone else stays, without its name), and the record is erased from our systems 30 days later.
- Close your account — runs the deactivation + 30-day grace + hard-delete cascade described above. The hard-delete is the only path that removes Conflict / Quick Repair / Growth / Trip / Daily / Joint-Onboarding content.
To ask for specific content to be erased while keeping your account, write to aoci@refairly.app. We handle such requests as data-protection law requires.
8.2 Preservation holds
We reserve the right to delay, suspend, or decline deletion or purge requests where we have a reasonable basis to believe that data preservation is necessary for pending or anticipated legal claims; fraud or abuse investigations; law-enforcement or regulatory requests; child-safety concerns; threats of harm; or compliance with legal obligations. Data subject to a preservation hold may be retained beyond normal retention periods until the hold is resolved.
8.3 Backups
Our database backups are kept for 7 days for disaster recovery and then deleted. They are used only to restore the Service. We do not edit backups, so deleted data can remain in them until they are deleted. If we ever have to restore one, we re-apply the account deletions and consent withdrawals recorded since it was taken, as far as our records let us find them.
9. Your Rights
Under GDPR, KVKK, UK GDPR, PIPEDA, Quebec Law 25 and other applicable data-protection laws, you have the following rights. To exercise any right, use in-app controls where available or contact aoci@refairly.app. Before we act on a request made by email, we confirm that it comes from you — normally by asking you to write from the email address on your account — and we may ask for more information where we have reasonable doubts about who is asking; we may refuse a request that is manifestly unfounded or excessive, as the law allows. We respond within the time the law sets: 30 days under KVKK, free of charge (if answering costs us more, at most the fee in the Kurul's tariff), one month under GDPR (extendable where the law allows). The ways to apply under KVKK are set out in the KVKK Aydınlatma Metni.
KVKK-specific rights (Turkey): under KVKK m.11 you have the right to learn whether your personal data is processed, request information about processing purposes and whether data is used in accordance with its purpose, learn the domestic or foreign third parties to whom your data is transferred, request correction of incomplete or inaccurate data, request deletion or destruction where processing conditions no longer exist, request notification of corrections or deletions to third parties to whom data was transferred, object to a result to your detriment produced solely by automated analysis of your data, and claim compensation for damage you suffer because your data was processed unlawfully.
Young people and parents: a young person with a teen account exercises these rights themselves, in the app or by writing to us; we may ask a parent to confirm identity. A parent exercises the rights over the record of a child who has no account. Where a parent's request would reveal more of a young person's entries than §10.3 gives her parents, we refuse it: those entries belong to the young person. Only what the law requires of us, or a binding order of a court or other competent authority, changes that (§10.12); we then share only what is required and, where the law allows, tell the young person.
9.1 Right to access & data portability
The app has a data export for every account (Settings → Download my data). The export includes, for the requesting user:
- Account profile (display name, email, language, country, city, bio, date of birth, gender) and your individual onboarding answers
- Each consent you gave or withdrew, and your acceptance of the legal documents, as it stands now: the version of its text, when you gave it and, if you did, when you withdrew it (a box you left unticked shows as declined)
- Subscription state + bonus token balance + period info
- Last 500 token-usage log rows (no prompt content)
- Last 500 notification-history rows
- Connection requests sent and received
- Joint Onboarding responses (your raw answers), relationship facts
- Daily Reflection answers and your Reflection Insights digest if any, your words for the day, messages you authored in the couple's chat
- Growth Plan responses (your raw answers); the goal items you wrote or last changed, your own goal items and those on the lists of the children you parent now, and the notes you added to goals, with the names of the files you added
- Trip plans you started (the wish), your own lines of things to plan around while they exist, and the trip chat messages you wrote, each with the app's record of what Refairly's plan tools did for it (§4.1)
- Your taste profile
- Conflict descriptions, your private session messages (including AI replies addressed to you), your shared summaries (any state), agreement signatures with your timestamp, check-in answers
- Quick Repair messages you sent
- Buzz messages you sent or received
- Buzz Call metadata where you were caller or callee (no intensity samples — those are not stored)
- Family: for each child you are a parent of, the record, your own observations, the expenses and pocket-money entries you created; chat messages you authored; your Refairly chat, both sides, the cards Refairly made for you and the notes it keeps about you; safety reports you filed
- For each person you look after in the app — a child, Mom or Dad, a pet — the record, its card and day, the notes you wrote, and for Mom or Dad the record of their answers (who recorded each — you, or our support when they wrote to us themselves — when, and which words)
- Your own day, with your medicines and their records and your partner's changes on it; on your partner's own day, and on a mother or father your partner looks after, only what you wrote there, and only while you are partners
- The Day: the whole day of each person you keep now — a child, Mom or Dad, a pet — items marked Parents only included, with their changes and every record on them, and on one you no longer keep only what you wrote there; the household's Home day, whole for every household you are or were a member of, and on one that went with a family to the other parent only what you wrote there
- For a caregiver account: the people you look after now — children, pets, a mother or father — their names, their keepers' names, since when and whether you run their day — and on their days the items you added and the records you made, never on an item marked Parents only and never their card or the record of a mother's or father's answers. A person whose link to you has ended is not in it
- Alarms you set or are on (for adults and young people alike): a reference to the item or goal each is on, when it rings, its label, whether it is on, whether you set it, your answer to it and whether it is on for you
- For a teen account: your own check-in entries, every theme shared about you, how many times the safety check showed you resources, your pocket-money entries, your goals and the notes you added, your own day without the items marked Parents only, the Home day items given to you and your own ticks on its other items, your own ticks on the day of a pet at home with you, your chat messages, and your Refairly chat with its cards
The export currently excludes AI-derived joint analyses (ComplaintAssessment per-user behavioral assessments; the AI-written 7-section Growth Plan reading; the Kitchen's AI-written proposals, apart from the agreed text of one you signed, which is in it with your signature; the AI-synthesized "About us"; AI replies in shared chat threads where you weren't the author; AI-generated Trip Plan content versions). It also excludes your relationship-check answers, an alarm's own heads-up (your own heads-up on an alarm is in it), the app's record of which reads each of Refairly's turns ran in your own Refairly chat (§4.7), the history of your consents with the IP address and device recorded with each, your notification settings, your devices and sign-in sessions, the facts about your relationship other than its type and whether you are expecting, and the files themselves — your profile picture, receipts, papers and the files on goals, which the export lists or marks but does not hold. These, and anything else we hold about you, can be provided upon written request to aoci@refairly.app.
9.2 Right to erasure (right to be forgotten)
- In-app or via web — Settings → Delete account; or visit refairly.app/delete-account and complete the email-verification flow (6-digit code, 15-min validity).
- By email — write to aoci@refairly.app.
On request: your display name is removed and your account is deactivated; your devices are signed out (a device may stay signed in for a short time, until its sign-in expires); your active relationship is ended — with children or a pet you keep together, by leaving the family to your partner, who carries it on (§8.1) — and your partner is told, and keeps any subscription their own account holds; your email address is kept during the grace period solely so the account can be restored. Underlying account data is retained in a restricted state for approximately 30 days, during which restoration is possible: signing in with Google or Apple, or registering again with the same email and confirming it from the email we send, restores the account; until that address is confirmed, the deletion goes on. A restore brings back what the request put out of reach — your sign-in, your consents as they stood, your bonus tokens and, once Apple or Google confirm it is still paid, your plan, your notification settings and inbox, and your place and messages in your chats, with your buzzes and calls; for a young person, her link to her record if it is still there and nobody has claimed it since, with her check-in entries and the notes she added to her goals. It does not bring back your display name, which you set again; a relationship ended by the request, a family it left to your partner and the chats left with it; a record only you kept; the alarm requests you had sent; or a young person's themes, safety-check records and claim codes (Deletion Policy §2). A password login is refused during the window; registering again with the same email brings the account back instead. There is no restore after the window, or for an account Refairly closed. After the grace period, an automatic daily process deletes your account. This removes, among other things: your account and sign-in records; your relationships, with their Conflict, Quick Repair, Growth and household-goal, Home day, Daily, Trip, Joint Onboarding, words for the day, Buzz and Buzz Call content; your subscription records, token usage, notifications and their settings, and device tokens; your Refairly chat, its cards and Refairly's notes about you, your own day, your own goal items with their notes and stored files, your observations about children, the messages you wrote in chats, your buzzes and calls, your stored profile pictures, and the receipts on the expenses that go with your account (a receipt on an expense that stays with the family stays with it). Exceptions are the records listed in §8 (consent, billing, safety and backup records); if we find something that was not deleted as described, we delete it when we find it. Child records you are a parent of stay with the other parent; one you were the only parent of is deleted when you ask to delete your account (unless a young person has her own account on it), and restoring the account does not bring it back; goals and notes you wrote on a child's list stay with that list, without your name. If you are Mom or Dad's son or daughter in the app, their record is deleted when you ask to delete your account, and restoring the account does not bring it back; the record of their answers stays (§8). A pet's record stays with the other keeper, or is deleted when you ask if you were its only keeper. A caregiver's links to children are deleted with her account; a parent's id is removed from the caregiver links that stay with the other parent (§10.13). A request you sent for an alarm that nobody answered is withdrawn when you ask to delete your account; at erasure your place on every alarm is deleted, and an alarm you set stays for the people on it, without your name (§2.10).
A teen deleting their own account removes at once their entries, the themes made from them, the safety-check records about them, their claim link and their place on alarms, and takes the notes she added to her goals off her parents' screens; at the end of the grace period those notes with their files, and her Refairly chat, are deleted. The child record itself, with her goals, stays with the parents as a record; restoring her account within the window links it to her again if it is still there and nobody has claimed it since, with her entries and those notes, but the themes and safety-check records do not come back. A parent cannot delete a teen's account for them; a parent can freeze it (§10.6).
Records kept after deletion. Your consent records (what you agreed to, the text version, when, and the IP address and device used) are kept after your account is deleted, to prove consent was given, for as long as we may need that proof under the law. Store notification records and the history of your plan are kept for billing, tax and fraud-prevention purposes, for as long as we may need them under the law; they then point to an account that no longer exists.
Important exception: we may retain data beyond the normal schedule where required by law, where needed for fraud prevention or safety, or where data falls under a preservation hold (§8.2).
9.3 Right to temporarily suspend your account
An adult can deactivate their account without deleting it (Settings → Deactivate account): this signs them out and pauses the account without deleting anything, and signing in again turns it back on. A teen or caregiver account has no deactivation; it can be deleted (§9.2).
9.4 Right to rectification
You can update your profile information at any time through the app. During multi-step flows (Conflict, Trip Planner, Growth Plan) you can submit revised content where the feature's state allows it. Parents can edit a child's record — its card and health list — and goals at any time. An observation added to the "Notes and papers" of a child, of Mom or Dad or of a pet is not changed or removed in the app once added; to have one corrected or removed, write to aoci@refairly.app (§8.1). Rectification of historical records may be limited where the record serves as a shared / signed dispute or agreement.
9.5 Right to object & restrict processing
- Push notifications: revoke at the OS level, toggle the master switch in app settings, configure quiet hours (they follow your phone's time zone), or opt out per-type where offered.
- Email: only transactional emails are sent (account confirmation, password reset, account-deletion verification, welcome email); these cannot be disabled because they are required for account security. No notification-mirror or marketing emails are sent.
- Analytics: off until you turn it on; you can turn it off again at any time in Settings → Usage analytics. Never on for a teen account.
- Consents: withdraw any of the consents in §2.3 by email, or in Settings where the app shows a switch for it (§2.3).
- Refairly's cards: dismiss any card; nothing changes unless you tap it.
- Refairly's notes about you: see every note, and delete any of them or all, in Settings → What Refairly remembers (§4.7).
- Alarms: decline a request, or turn an alarm off for yourself; it then does not ring you (§2.10).
- Legitimate-interest processing: you have the right to object to any processing based on legitimate interest. Contact aoci@refairly.app; we will assess whether our legitimate interests override your rights and freedoms in your specific situation.
- Restriction: you can ask us to restrict the processing of your data — for example while you contest its accuracy, or while we assess an objection — by writing to aoci@refairly.app; we then keep it but do not otherwise use it, except as the law allows.
9.6 Right regarding automated decisions
See §4.3.
9.7 Right to lodge a complaint
- Turkey: Kişisel Verileri Koruma Kurumu — www.kvkk.gov.tr
- EU/EEA: your local Data Protection Authority
- UK: Information Commissioner's Office (ICO) — ico.org.uk
- Canada: Office of the Privacy Commissioner of Canada — priv.gc.ca; Quebec: Commission d'accès à l'information — cai.gouv.qc.ca
9.8 Jurisdiction-specific rights
If a mandatory data-protection law in your jurisdiction grants you additional rights or imposes stricter requirements than those described in this Privacy Policy, those mandatory rules apply to the extent required by law. Nothing in this Privacy Policy is intended to limit any rights that cannot be limited under applicable mandatory law.
10. Children, Young People & Parents
10.1 Ages
- Adult accounts are for people aged 18 and older. The 18+ requirement is a contractual eligibility condition you affirm when accepting the Terms of Service at sign-up; we additionally ask for an optional date of birth during profile onboarding. We do not perform document-based age verification for adults. If we learn that an under-18 person has created an adult account, we may suspend or terminate the account and delete associated personal data, subject to lawful retention requirements.
- Teen accounts are for young people aged 13 to 17, and can only be created through a parent (§10.2). Nothing changes by itself when the young person turns 18: the account stays a teen account until she deletes it (§9.2); her parents keep seeing what §10.3 lists and can still pause it (§10.6), but cannot delete it. She cannot leave the family link herself; she can delete her account at any time in Settings.
- Children under 13 do not have an account. A parent may keep a record about them (§2.1, "Children as records"); the child does not use the Service.
- Mom or Dad has no account. Their son or daughter keeps a record about them (§10.15); they do not use the Service themselves.
- Caregiver accounts are for adults aged 18 and older whom a family asks to help look after someone — a child, a pet, or a mother or father (§10.13). A caregiver is not a keeper of the person in the app.
If you believe a minor has created an adult account, or a teen account was created without a parent's involvement, contact aoci@refairly.app.
10.2 How a teen account comes to exist — and why this is our parental-consent mechanism
A teen account cannot be created from a sign-up screen. It exists only when:
- A parent who already uses the Service has added the child to their family with the child's date of birth, after giving the children's-data consent in §2.3.
- The Service has computed from that date of birth that the child is at least 13 (
MinimumClaimAge, enforced on the server at the moment a code or link is issued). - The parent, on their own phone, asks for a one-time claim code (a hashed 6-digit code shown as a QR, valid for 15 minutes, one live code per child), or a one-time link to send her (hashed, valid for 24 hours, one live link per child).
- The young person redeems that code or link on their own device, creates their sign-in (an email address, which may be a parent's, a password and a display name), reads the teen contract screen in plain language — what parents see, what stays theirs, and the safety rule — and decides whether their check-ins may be turned into themes.
The parent's act in steps 1–3 is deliberate and recorded; it is the authorization of the person with parental responsibility that GDPR Art. 8, Quebec Law 25 and Turkish law require, and we rely on it in every country regardless of that country's own age of digital consent (§10.9). Both parents of the child are notified when the account is claimed. The young person's consent in step 4 is recorded with the time, IP address and device, and the account is not created if it cannot be recorded.
10.3 What parents see, and what stays the young person's
The split is enforced on the server, not by the app:
| Parents see | Parents do not see |
|---|---|
| The child's record (name, birth date, gender, the card and the observations parents wrote themselves), her goals with their notes (including any the safety check flagged), photos and files, and her whole Day — the items they set, the items she adds herself (medicine too) and every tick she makes — and the alarms on her goals and her Day, with whom each rings and how each person answered (§2.10) | — (nothing on her Day is kept from them) |
| The mood numbers she picks in her check-in (her screen says so when she answers), and her words only while "share my plain words" is on — what she wrote from the moment she turned it on; answers the safety check flagged are shared like the rest while it is on, and her screen says so first | Her words while sharing is off, and anything she wrote before she turned it on. Turning sharing off hides every answer from them again |
| Weekly themes made from the teen's check-ins, with the teen's consent — short phrases like "wants calmer evenings", one per area of life; home themes are phrased as needs, not accusations | Quotes, details, or which entry a theme came from |
| Pocket-money entries, the family group chat, the teen's chat with all her parents, and the plan and chat of any trip she is on, even one they are not on | Each parent's chat of the two of them with her (the other parent does not see it); her chat with her caregiver |
| That a safety signal about school or outside life occurred — as a contentless "please check in with them" (§10.5) | Any safety signal about home; the AI's description of any signal |
| That the account was claimed, paused or deleted | The teen's password; the teen's email once set |
Both parents of a child see the same things. The young person sees every theme that was shared about them, how many entries they wrote, and how many times the safety check showed them resources, on their transparency log. Pausing or revoking a teen account does not open her words to a parent (§10.6). A caregiver the parents choose sees less than either of them, and §10.13 sets out exactly what.
10.4 The young person's consent
The consent a young person gives at claim time — changeable later in their settings — governs whether their entries are distilled into themes and whether the AI replies in chats they are in. Declining is a real answer: entries are still saved, and no themes are made. Withdrawing it later deletes the themes already made, and the copies of them posted to her parents. The same consent lets her own Refairly make cards she taps, and lets a parent's Refairly read a chat she is in; while it is withdrawn, neither happens. Her Kitchen chat and table rest on a consent of their own, which she gives on her own sheet (§2.3). A parent cannot give or withdraw either consent on the young person's behalf.
10.5 The safety check in family surfaces
What it is. Each teen check-in entry, each message a young person writes to her own Refairly chat, each note she adds to a goal, each message she herself sends to Refairly in a chat — with @refairly, or in a trip plan's chat while Refairly's question before a build is open — or asks it to help her say, and — before it is stored — what she writes in her Kitchen chat or at her table, is evaluated by the AI for signs that a young person may be in danger — to themselves or from others. A message someone else sends to Refairly in a chat with her is not checked, and neither are chat messages that are not sent to Refairly, including when a member's own Refairly reads the chat (§4.7); the weekly check-in is where the check runs every week. This runs whether or not the young person consented to themes and replies, because a safety net is not something consent switches off; where consent is withdrawn we rely on the protection of vital interests (§3).
What it can do — and all it can do.
- Show the young person help resources. General helplines, which may not be right for every country, and the reminder that the app cannot help in real time. In her Kitchen chat or at her table, a signal also pauses the chat or the table and keeps her words for her alone.
- Ask a parent to check in — only for signals about school or outside life. The notification carries no content: "please check in with them". It does not say what was written.
- Leave parents out of a signal about home. A disclosure about home is about the people who would receive the alert, so the young person is shown resources and nothing else happens. If the AI cannot tell which area of life a signal belongs to, we treat it as home — the non-notifying default.
What it is not. It is an automated aid, not a clinical instrument and not a promise. It will miss situations and it will sometimes flag situations that are not dangerous. At times — for example during an outage — it may not run at all, and nobody at Refairly watches its results in real time. It does not contact emergency services, does not read a chat a young person is not in, does not run on the couple's own conversations, and is used only for the actions above. If a young person is in danger, contact emergency services; do not rely on this check.
What we keep. A record of each event: the area of life, the AI's short description of the signal, whether resources were shown and whether parents were asked to check in. The description is blanked after 90 days and the record deleted after 365 days; the mark in her own Refairly chat that help resources took the place of a reply is removed from that message's record after 365 days too. The young person can see how many times the check showed them resources; the descriptions are seen only by us, on a report or a lawful request. While a young person shares her plain words, her parents read a flagged answer like any other (§10.3). A note she adds to a goal stays on the goal's page, which her parents read, whether or not the check flagged it.
10.6 Pausing, revoking and deleting
A parent can pause (freeze) a teen account. While paused the young person cannot use it beyond reading — her goals and the goal items of her household, her Day and the Home day, and her pocket money — and her phone is given no alarms to ring (§2.10); pausing, revoking or deleting does not open more of her entries to a parent. A parent can resume a paused account. A young person can delete their own account at any time (§9.2). A child record with a teen account attached cannot be deleted by a parent — pausing is the strongest step a parent can take — and once the young person's account is gone, the record can be deleted like any other.
10.7 Billing
A teen account cannot buy anything, sees no store, and is not charged. AI turns in a chat with a young person, her own Refairly chat, her Kitchen chat and table, and the distillation of their check-ins, are paid from a parent's pool; if no parent in the chat has an active pool, the AI does not take a turn in the chat. The safety check runs regardless.
10.8 A young person's rights
A young person with a teen account can export their data, delete their account, withdraw their consent, and ask for human review of a theme or a safety-check event, in the app or by writing to aoci@refairly.app. We may ask a parent to confirm identity, but we will not disclose the young person's entries to the parent in doing so.
10.9 Country rules
- EU/EEA (GDPR Art. 8): member states set the age of digital consent between 13 and 16. Because every teen account is created through a parent's deliberate act (§10.2), the parental authorization Art. 8 requires exists for every teen account, whatever the member state.
- Quebec (Law 25): a minor under 14 needs a parent's consent. The parent-initiated claim provides it; a 13-year-old in Quebec cannot obtain an account any other way. Quebec users may contact the person in charge of the protection of personal information at aoci@refairly.app.
- Turkey (KVKK and the Civil Code): a minor's consent is given with the involvement of the person holding parental authority; the parent-initiated claim is that involvement, and the young person's own explicit consent is recorded for the sensitive processing under KVKK m.6.
- United Kingdom: the age of digital consent is 13; the parent's involvement is additional to it.
10.10 The Service is not appropriate for certain users
You should not use Refairly if:
- You are in a relationship that involves abuse, coercion, threats, stalking, financial control, or any form of intimate-partner violence. A shared two-person AI tool can be weaponized in such relationships to surveil, pressure, gaslight, or control the other partner. If that is happening, end the relationship or close the Kitchen chat, report it, or close your account (§10.11).
- You or your partner is in acute mental-health distress, active suicidal ideation, active self-harm behavior, or active substance-use crisis that requires clinical attention. Refairly is not a clinical service.
- You are subject to a restraining order, no-contact order, protection order, or any other court order restricting contact with the person you would invite as your partner or with a child you would add as a parent. Using Refairly to contact such a person may violate the order.
- You need immediate emergency intervention of any kind. Contact local emergency services.
- You are using Refairly as a substitute for professional therapy, couples counseling, mediation by a licensed arabulucu, legal advice, or medical care. The AI cannot replace any of these.
10.11 If you experience harm through Refairly
If your partner or anyone in a family chat is using the Service in a way that you experience as harassment, coercion, abuse, control, or psychological pressure, you may end the relationship (with children, by leaving the family to your partner, which does not need anyone's agreement) or close a Kitchen chat in-app at any time (the couple's and the family's chats are left by ending the relationship or leaving the family), report the message or person with the report control in the app's chats or to aoci@refairly.app, close your account, and — most importantly — contact local authorities or appropriate helplines if you are in danger. Reports are reviewed by a person. We don't promise a response time, and Refairly cannot act in an emergency. Refairly is not an emergency service and cannot intervene in real-world safety situations.
10.12 Cooperation with law enforcement and lawful requests
We cooperate with law-enforcement requests, court orders, and other lawful requests to the extent required by applicable law. This may include preserving or producing user records, including content, metadata, and audit logs.
10.13 Caregivers — an adult a family asks to help look after someone
Some of the people who look after a child, a pet or a mother or father are not family: an au pair, a babysitter, a dog walker, a neighbour on a Tuesday. Refairly lets the keepers give such an adult what there is about the person, except what they keep to themselves. What follows is written for a child, as most caregivers look after one; it holds the same for a pet and for a mother or father, with the differences set out below.
How it starts. A parent asks, on their own phone, for a six-digit code and shows it in person, or sends a link instead. Both are hashed and single-use; a code expires in fifteen minutes and a link in 24 hours, and one of each is live for a child at a time. Anyone who has a link can use it, once, so it is meant only for the person it is sent to. The adult enters the code, or opens the link, on their own device, sets their own email address and password, and gives their date of birth; the Service refuses the account if that date makes them under 18. The account exists but holds nothing until they confirm their email — only then is the child linked to them, their chat with the parents opened, and both parents notified. We do not verify the date of birth with documents; the parent knows who they gave the code or link to, and that is the act that matters. A caregiver who already has an account adds another family the same way, from inside her app: signed in, she enters the code a parent shows her, or opens the link a parent sent her. Her email is already confirmed, so the child is linked to her at once and both parents are notified. A code meant for a young person's own account does not work there.
A caregiver the family already has. A keeper can also give another person they keep — a child, a pet, or their mother or father — to someone who looks after, or looked after, anyone else they keep, without a new code. That person's page lists them — those who look after someone now first, then those who "looked after before" — and the keeper chooses one and confirms. She is not asked first: the person is in her app at once, she is told who gave her whom, and the other keepers are told too. The list holds only the caregivers of the people that keeper keeps whose account is open and confirmed. It leaves out a caregiver who left the family herself, unless a parent has given her a child since; a new code or link still works for her.
What a caregiver sees: for each child a parent has given them — the display name, the age in years, and the child's card (the parents' note, what the child eats, the emergency contact and the health list); the child's Day for today — every item except those marked Parents only, a young person's own items (medicine too) included, with every tick and medicine record on them — to which, while a parent leaves that switched on, the caregiver can also add items for today (not a medicine), change or skip today's items and tick them (a young person's own items stay hers; the parents are told when she skips or changes something today); the course logs of the medicines she sees; the alarms on those Day items and on the child's goals — she can set one for herself on any of them, answer one she is asked to accept, and, while a parent lets her change the Day, set one on a parent's item that asks the others who see it (§2.10); the child's papers, except those the parents keep to themselves; the child's goals that are not Parents only, with their notes, photos and files, and the child's pocket-money entries, both to read and not to change; the chat with the parents; when a parent opens it, a chat with a young person she looks after; and her notifications about these children, in her in-app inbox. A caregiver given Mom or Dad sees their card, day and goal items — and their health list, medicines and papers once they have agreed in person — and not notes, anything marked Parents only, the record of their answers or who else helps. Their medicine reminders reach the caregiver who runs their day while the family's plan is active. A caregiver given a pet sees its card, its day and its papers, except those its keepers keep to themselves.
What a caregiver does not see: a young person's check-in entries, words, themes or moods; a parent's observations about any child; anything the parents keep to themselves; the family's plan, agreements, expenses or budgets, or the household money a pocket-money top-up appears in; the goal table, the household's goal items and the parents' own, and the Home day; anything between the parents; the family pulse; any other child, in that family or another, unless the parents give them that child too. The server has no route that returns any of it to a caregiver account, and the couple surfaces answer "forbidden" to it exactly as they do to a teen account. Caregivers have no Refairly.
The chats. One chat per family a caregiver looks after: the parents and her, about the children she was given. Refairly does not take a turn in it — writing @refairly there does nothing, and the app does not offer the AI. No child or young person is a member. A parent can also open, from the caregiver's screen, a chat of a young person and her caregiver, the two of them: her parents are not in it and do not read it, Refairly is not in it, and a parent can close it at any time. A caregiver's chat about Mom or Dad is between their son or daughter and the caregiver; when a couple ends it stays with them.
Taking it back. Either parent can remove a caregiver from a child at any moment, without the other parent's agreement and without notice to the caregiver beyond the child disappearing from their screen; it also takes the caregiver off the alarms on that child's Day, and her notifications about that child leave her inbox (§8); taking a young person back also closes her chat with that caregiver; when the last child of that family is taken back, the caregiver's chat with the parents closes. The caregiver can leave a family the same way, at any time. A link that ends this way is kept as history, marked as removed: it shows nothing to the caregiver, and it is what lets the parents' list offer her again as "looked after before" (a caregiver who left herself is not offered). Deleting the child's record deletes every link to the child, past ones too; deleting the caregiver's account removes every link she had; when a parent's account is deleted, their id is removed from the links that stay with the other parent (§8). None of this touches the child's record, which stays with the parents.
Lawful basis. The parents disclose the child's card, including the health list, the child's Day, including its medicine records, and the child's papers, goals and pocket money, on the basis of the children's-data consent they gave before adding the child (§2.3), and they choose the recipient themselves. We process it to deliver the feature they invoked. A caregiver is a recipient of that data, not a controller of it: they cannot change the record or download the child's card, and lose access the moment a parent removes them. Her own data download holds what is hers: who she looks after now, and the items she added and the records she made on their days, never on an item marked Parents only (§9.1). Once her link to a child ends, that child is no longer in it.
10.14 A child's papers
Parents can keep files about a child on the record: a health report, a letter from school, a photo. Images, PDF and Word documents, up to 15 MB each.
Who may open one is chosen per file, by the parent who adds it, because a vaccination record and a custody letter are not the same kind of thing:
| The choice | Who can open it |
|---|---|
| The family (the default) | Both parents, the young person herself once she has her own account — a report about her body is hers to read — and the caregivers the parents gave this child (§10.13), for the allergy report a sitter may have to show at a hospital desk |
| Parents only | The child's parents. It does not appear on a young person's screen or a caregiver's, and nothing there says a file like it exists |
How it is held. The file is stored privately on AWS S3 (encrypted at rest) and is not public: every read is a link signed for ten minutes, issued to a caller the file's own setting admits.
Refairly can read a paper when it needs one to answer (§4.7): it lists the papers the person who asks may open, and reads one — an image (JPEG, PNG or WebP) or a PDF; it does not read Word or HEIC files — only if that person may open it, and in a chat only if every member may. A young person's Refairly does not see papers kept to her parents, or their names. The file is sent to Anthropic for that answer; a photo is sent scaled down and without the details a phone stores in it, such as where it was taken. The photos and PDFs on a goal's notes are read the same way, for someone who may see the goal. Papers are not sent to the AI in any other surface.
Who may change it. Adding, re-aiming and removing a paper are parents' acts, and either parent can do all three — re-aiming and removing also by a card from Refairly that the parent taps. A young person reads what is meant for her but does not delete what her parents keep; a caregiver can do neither. Removing a paper deletes the stored file with it.
When it goes. A child's papers go with the child's record when a parent deletes it, and with the last parent's account at erasure — the rows and the stored files both (§8, §9.2).
Lawful basis. The children's-data consent in §2.3, given before the first child was added, whose text names the papers and health reports specifically, and Refairly reading them. Withdrawing it stops new papers; the ones already there stay for the other parent until the record is deleted.
10.15 Mom or Dad, looked after by the family
Some families look after a mother or father — of one of the partners, or of someone on their own. Refairly lets that parent's son or daughter keep a record about them — their card, their day, their goal items, notes and papers — and share it with the people who help, much as parents keep a child's record. They have no account of their own.
Before adding them. Their son or daughter confirms a statement: that they are their son or daughter; that they will tell them what is kept and who sees it; that they will record their agreement only when they give it in person; and that until then they will leave health details out of their day, goals, card and notes. The statement is not a consent to their data and is not the legal basis for it ("Lawful basis" below). Taking it back in Settings ("Mom or Dad's details") stops new notes, papers and files about them.
What waits for them. Their health list, medicines and papers stay locked until they agree in person, and Refairly reads nothing of their record until they agree to that as well: two answers on one page. Their son or daughter shows them the words the app has for them, or reads them aloud, and records each answer only when they understand it and say yes; the record keeps who recorded it, when, and which words they were shown. If they cannot understand what they would agree to, their health list, medicines and papers stay locked and Refairly does not read their record; a guardian's consent on their behalf is not supported.
Who sees what. Their son or daughter sees all of it. That person's partner (not a co-parent) sees them through the partner link while the couple lasts, may tick their day, and gets their medicine reminders only after choosing "Ring me too". Notes and items marked Parents only are for the two partners. Someone the family asks to help look after them sees their card, day and goal items — and their health list, medicines and papers once they have agreed — and not notes, anything marked Parents only, the record of their answers or who else helps. A young person at home sees their goal items in the household's table unless they are marked Parents only, and nothing else about them. Their medicine reminders go to their son or daughter and, while the family's plan is active, to the caregiver who runs their day; on a locked screen they say only "Time for {name}'s medicine".
Refairly. Once they have agreed to it, Refairly reads what the person who asks may see of them — for their son or daughter or that person's partner, and in a chat only where every member is one of those two — to answer and to suggest cards that person taps; what it needs is sent to Anthropic, in the USA or other countries (§5.2, §6). Before that it reads nothing of their record; what someone writes about them in a chat is sent as that person's words.
When a couple ends. They stay with their own son or daughter, even when that is the one who leaves, and the other partner's access ends at once. A caregiver's chat about them belongs to their son or daughter and stays with them. Expenses recorded for them stay in the household's money, which goes to whoever keeps the family (§8.1); their name on those lines shows only to the people who can see them.
Changing their mind. They can change their mind about either answer at any time, through their son or daughter or by writing to us. Their son or daughter records it in the app. For the first answer, their health list, medicines and papers are deleted with their records, and so is the past as a medicine of one of their medicines changed into another kind (on an older item whose change of kind was not recorded, that past is erased 30 days after the item is removed); for the second, Refairly reads nothing new of their record, and its earlier replies stay in the chat's recent history.
Their rights. They have the rights in §9. They can use them through their son or daughter, or directly by writing to aoci@refairly.app; we confirm who they are with them directly, never through their son or daughter unless they ask us to. When they ask us to take back one of their answers, we record it for them, and the record of their answers shows that our support recorded it. If they object to the record, we delete it. Either way we tell their son or daughter.
Lawful basis. Their record without the health list, medicines and papers rests on legitimate interests (§3): our interest in providing the feature their family uses, and their family's in looking after them, balanced by keeping no health list, medicine or paper and sending nothing of their record to the AI before they agree, by their son or daughter telling them what is kept, and by asking their son or daughter to leave health details out of the rest. Their health list, medicines and papers, and Refairly reading what is kept about them, rest on their explicit consent, given in person, one answer each (Art. 9(2)(a); KVKK m.6/3-a).
What we keep. Their record goes when their son or daughter deletes it, or when that person asks to delete their account, and a deleted record is erased from our systems 30 days later (§8, §9.2). The record of their answers — who recorded each, when and which words, with no health data — is kept like other consent records (§8).
11. Cookies, Tracking & Mobile SDKs
11.1 Mobile application
The Refairly mobile app does not use browser cookies. The app includes the third-party SDKs listed in §5 (Firebase Analytics, Firebase Crashlytics, Firebase Cloud Messaging, Google OAuth, Apple Sign-In, Apple StoreKit, Google Play Billing), WebRTC for the voice in Buzz Calls (§2.11), the store's own rating sheet (Apple's StoreKit review request on iPhone, Google Play In-App Review on Android), and Expo's update service (expo-updates; Expo is 650 Industries, Inc., in the United States): each time the app starts, it asks Expo's servers whether a newer version of the app's own code is ready and downloads it, and Expo receives the phone's IP address, its platform, the app's version and a random identifier for this installation that the app keeps on the phone — nothing you wrote in the app. Like crash reporting and push delivery, this check cannot be switched off while the app is installed. Analytics is off until you turn it on, and never on for a teen account (§5.4). Your answer is kept on the phone, and signing out clears it, so the next account on that phone starts with analytics off.
The rating sheet. Now and then, after something has gone well in the app, an adult account may be shown Apple's or Google's own rating sheet. Whether and when it appears is up to the store, and a rating or review you leave there goes to Apple or Google under their terms; we are not told whether you rated. To decide when to ask, the app keeps a small record on the phone for each account — when it first saw you, on how many days you used it, how many of Refairly's cards you tapped, and when it last asked — and sends none of it to our servers. Where analytics is on, the app notes that the sheet was asked for (§5.4). Teen and caregiver accounts are not asked. Crash reporting and push delivery cannot be disabled while the app is installed (notifications themselves can be disabled at the OS or in-app level).
11.2 Website (refairly.app)
The website uses no advertising or analytics cookies or trackers and keeps no one signed in; the only thing it stores is a short note in your browser's session storage, for the open tab, so that a download link sends a phone to the app store only once.
12. Subscriptions & Payments
Subscriptions and consumable token packs are purchased through Apple App Store or Google Play. We do not process payments directly. We receive only transaction identifiers and subscription status.
The welcome week. Every new household starts with seven days in which everything is open, the AI included, on one allowance the whole household shares; there is no plan to choose and no payment step when you sign up, and no partner needs to connect first. The week starts with the household's first adult account: a partner who connects shares the days left, and a household that has ever had a plan has no welcome week. Two days before it ends, each adult in the household is told once, at 20:00 on their own phone's clock. Buying a plan during the week ends it, and the plan takes over; a token pack bought during the week adds to its allowance, and what is left of the pack waits for a plan started after it. When the week ends without a plan, the app shows the plans before anything else; your records are kept, and no further week without the AI follows. Young people and caregivers in the household use the week as the family does, and are not told about it.
A Refairly subscription belongs to the account that buys it. In a couple, one partner subscribes and both partners receive the entitlement; a user without a partner can subscribe for themselves. Token-pack bonuses (they do not expire) are added to the buyer's own subscription if they have one, otherwise to their partner's. Both partners can see which of them pays, each subscription's plan and renewal date, and — before it leaves the pool — a subscription that won't renew; they do not see each other's billing details. A partner who pays into the pool can see each household member's share of the period's use, as a percentage — not what anyone asked or wrote. The family features draw on the same pool: a teen account has no subscription of its own, cannot purchase, and is not charged.
Apple and Google tell us about renewals and refunds, and we check that each notification really comes from them before we update your plan.
For refund requests, contact Apple or Google through their respective platforms. Subscription cancellation is handled through your device's store settings. When Apple or Google refunds a token pack, the bonus tokens it added are deducted from the pool, down to zero at most; tokens already used are not taken back from anyone.
Detailed tier features, pricing, and current offers are displayed in the in-app paywall at the point of purchase.
13. Data-Breach Notification
13.1 GDPR (EU/EEA)
In the event of a personal-data breach likely to result in a risk to the rights and freedoms of EU/EEA or UK users, we will notify the relevant supervisory authority within 72 hours of becoming aware (GDPR Art. 33). If the breach is likely to result in a high risk to individuals, we will also notify affected users without undue delay (GDPR Art. 34).
13.2 KVKK (Turkey)
We will notify the KVKK Kurulu without delay and at the latest within 72 hours of learning of a breach (or, if later, with the reasons for the delay), and the affected data subjects as soon as reasonably possible once we know who they are, in accordance with KVKK m.12/5 and the Kurul's decision 2019/10.
13.3 Canada and Quebec
We will notify the Commission d'accès à l'information and the affected persons of any confidentiality incident presenting a risk of serious injury, as Law 25 requires, and the Office of the Privacy Commissioner of Canada where PIPEDA requires it, and we keep a register of incidents.
13.4 General
Not every security incident constitutes a notifiable breach. We assess each incident on a case-by-case basis considering the nature, scope, context, and likely consequences. We document all incidents and our assessment of notification obligations. Where the law requires us to tell affected people about a breach, we tell young people directly and, where appropriate, their parents — only what is needed, and without revealing words the young person keeps from them.
14. Changes to This Policy
We may update this Privacy Policy from time to time. We aim to notify you of material changes through one or more of:
- Asking you to accept the updated documents in the app before continuing
- An in-app notification
- An email to your registered email address
- Updating the "Last updated" date at the top of this page
Material changes take effect on the date shown at the top. For material changes we tell you in the app before they take effect and, where the law requires, ask for your agreement.
15. Contact Us
Ali Onur Can İrey Data Controller & Developer of Refairly; person in charge of the protection of personal information (Quebec Law 25)
Email: aoci@refairly.app Website: https://refairly.app
For privacy inquiries, data-access requests, erasure requests, or complaints, please email us. We respond within the time the law sets: 30 days under KVKK, one month under GDPR (extendable where the law allows).
© 2026 Refairly — Ali Onur Can İrey