KGB BABUSHKA

PRIVACY POLICY

Version: 2.0 · Last updated: 8 October 2026

The short version

The rest of this page is the detail.

1. Who we are

Pungent Ltd makes KGB Babushka and decides how your data is used. In data-protection language we are the controller.

Company Pungent Ltd, registered in England and Wales, company number 17117616
Registered office 167-169 Great Portland Street, London, W1W 5PF, United Kingdom
Email info@kgbbabushka.com, for privacy questions and requests. It reaches the person who decides.
UK ICO registration registration in progress (application ref C2053707)
EU representative Being appointed. We will name our representative in the EU (GDPR Art. 27) here, with a postal address and email, before the app is available on the App Store in the EU. Until then, people in the EU, including anyone testing a pre-release version, can write to us directly at info@kgbbabushka.com.
Person in charge of the protection of personal information (Quebec) The Director, Pungent Ltd · info@kgbbabushka.com · the registered office above
Data protection officer We are not required to have one.

This policy covers the KGB Babushka iPhone app and our website, kgbbabushka.com.

2. Age: 13 and over; the AI helpers are for adults

3. What we collect, and why

3.1 Your account

What Why
A random account ID, made by the app when you first open it. The bottom of Settings shows its first eight characters, so you can quote it if you write to us. So your progress and settings are yours and stay yours
Whether you are an adult or a teenager, and the record of a one-time correction (§2) To apply the age rules in §2
A security key for the app on your iPhone (Apple's App Attest): the app makes it when you first open it, and our server keeps its public half and Apple's confirmation. Where we use Apple's DeviceCheck, Apple also keeps two bits per iPhone for us, a count of how many new accounts that iPhone has started this month, which survives a reinstall and says nothing about who you are; our server keeps that count with the key. So our server can tell that a request to an AI helper comes from the genuine app on a real iPhone, and so that deleting and reinstalling the app does not reset the free daily allowance
Your App Store country, and the country your internet connection comes from, which Cloudflare works out from your internet address and passes to our server when you use an AI helper (never your position within that country) To apply the rules of where you are: which privacy defaults apply, which helpers are open in your country, and which helpline page and emergency number the crisis reply gives

We don't ask for your email, phone number or a password. There are no purchases in version 1.0, so we hold no payment or purchase records.

3.2 The guest book

The guest book is the book you fill in when you arrive in the village. You decide what to write.

3.3 What you do in the village

What Why
Progress: lessons started and finished, quiz and snack answers, time in a lesson, game results, the ideas you have met and kept, UP and the other in-game currencies, your rank To run the game and show you your own progress
Activity: what you open, what you skip, how long you stay on a lesson or reel, and the topics you choose in the feed's filter To choose what to show you next. If you talk with Babushka, she also sees a short summary of what you have done lately. Settings → Privacy & account → What Babushka may know → "Let me get to know you" turns this off. The default follows the rule in §6.1.
Notebook notes Stored so they sync between your devices. They are encrypted on your phone first, with a key kept in your iCloud Keychain, so that copy is scrambled text to us. A readable copy of each note you haven't marked Private is kept for Babushka only after you have said yes to her and then "Let her" (§5).
Reflections you write in lessons You choose where they live (Settings → Privacy & account → Where your reflections are kept): encrypted in the cloud with a key in your iCloud Keychain, on this phone only, or not saved at all
The Council: your questions and its answers Kept in your Ledger in the app, so you can read them again
Theatre takes: recordings you make in the Theatre (in version 1.0, reading one of the Great Speeches aloud) Your takes stay on your phone unless you turn on Settings → Recordings → Save takes to my account; then each take you record from then on is also stored with your account, and you can turn it off at any time. Speech-to-text runs on your phone.
Your time zone, and when you last opened the app, only after you say yes to Babushka's card So her daily thought reaches you at the hour you chose in your guest book, in your own time zone, and only while you are using the app (she writes it only for people who opened the app in the last 7 days)
Reports you send about one of Babushka's replies, with the part of the conversation you choose to include So a person can read what went wrong and fix it
Reports you send about a Council answer. If you report a Council answer, we keep it with your report: your question, the figure's name, the answer you saw, and anything you add. A person on our small team reads it. It is included when you export your data. So a person can read what went wrong and fix it
Emails you send us To answer you

Under 13, only your progress leaves your phone (and the few things §13 lists); the rest of this table stays on it, or is never collected (§13).

3.4 What you tell Babushka, and the other AI helpers

See §4 and §5.

3.5 Usage counts and crash reports

See §6.

3.6 What we don't collect

4. The AI helpers

Three parts of the village run on AI from other companies. In version 1.0 two of them are open, Babushka and the Council; the Theatre's voice rooms are closed, and the Theatre rows below say how they work once they open. Each one stays off until you agree. The first time you open that room, a card names the company and says what goes to it. Each card has its own yes, and nothing is sent until you tap it. "Not now" keeps only that room closed; the rest of the village works as before. If we ask again on Babushka's card (see "If we change an AI company" below) and you choose "Not now", her replies and her overnight notes pause until you say yes; nothing is deleted. Every answer, a yes or a "Not now", is recorded with the version of the words you were shown. The AI helpers are for adults (18+).

You can turn each one off at any time: Settings → Privacy & account → What Babushka may know. Turning one off closes that room, and nothing more is sent to that company.

Helper Company What it receives What it does with it
Babushka (the Kitchen) xAI (Grok), USA What you type or say to her (your phone turns speech into text first); your recent conversation; her notes about you; the lessons and courses you have finished; if "Let me get to know you" is on (§6.1), a short summary of what you've done in the village lately and what you are learning; your Notebook notes, only as §5 says. Her card names each of these before you agree, and new sources mean a new card. With each message xAI also gets a scrambled code for your conversation, so that it can reuse its own work between messages; it carries no name, and xAI cannot turn it back into your account ID. Writes her replies. Checks each message and each reply for safety (§15).
A photo you show Babushka xAI (Grok), USA The one photo you pick, after your phone has removed where and when it was taken and every other detail stored in the file, and made it smaller. xAI fetches it through a short-lived link to our storage, which carries neither your account ID nor your name. Covered by your yes on Babushka's card; you can switch photos alone off in Settings → Babushka → What she may do. Checks it for safety first: nudity, violence and anything that may show a child at risk are refused, and she is never shown them. Then helps her see it, to answer you. We delete it as soon as she has answered; we keep no copy, and nothing about it is written into your conversation except your own words.
Your alias suggestions (the guest book's last page) xAI (Grok), USA A short summary of what you wrote in the guest book, and any name you type yourself Suggests three agent names, and checks a name you typed. Only if you are 18 or over and have agreed to Babushka's card.
The Theatre (closed in version 1.0; adults only, with your yes): the Scenes, the Minute, and notes on your takes OpenAI, USA, for the voice; xAI, USA, for the notes In the Scenes your voice is streamed live to OpenAI (Realtime API), whose AI plays the other person; OpenAI does not train on it and keeps abuse-monitoring logs for up to 30 days. When a take or scene is judged, the transcript (text, never audio) goes to xAI, which deletes it under zero data retention and does not train on it. Google receives nothing from the Theatre. On its way to OpenAI, your voice passes through Cloudflare, which relays it and keeps no copy. OpenAI plays the other person in the scene. xAI writes the notes.
The Council OpenAI, USA The problem you describe, and your conversation with the Council Writes the Council's answers

Babushka does not search the web in this version. She answers from what she already knows, and she cannot open a link you paste.

Photos you show Babushka are for adults. Your yes on her card covers them, and you can switch photos alone off, and back on, at any time in Settings → Babushka → What she may do.

Photos and child abuse images. A photo that our safety check reads as possibly showing someone under 18 in a sexual way is refused and deleted at once. We never keep a refused picture, and no one on our team ever looks at one.

When our check reads a photo as clearly showing a child in what US law defines as child sexual abuse material (sexually explicit conduct, or a lascivious exhibition of the body: 18 U.S.C. §2256), photos close for that account, a person on our team decides from the record alone, and we report it to the National Center for Missing & Exploited Children (NCMEC), the US clearinghouse, which passes reports to the police where the person is. These are the only photos we report. A drawing, not a photograph, that our check reads as clearly sexualising a child is refused and photos close for that account, but it is not reported. A photo of a child with nothing sexual in it, such as a bath or a rash, is refused gently, with a suggestion to show a worry to a doctor or a pharmacist; photos stay open, nothing is reported, and no record says why it was refused. The report says what we hold: your account ID, the time and your App Store country. It never includes the picture, because we have not kept it. What a report contains is kept for a year, because US law treats a report as a request to keep its contents, even if the account is deleted in that year; then only that record is kept, locked away, and everything else is deleted as usual. If the account is deleted before we have decided, we keep only the flag (never a picture), locked away, until a person decides: it is then reported, or deleted. xAI, the company whose AI checks photos, has its own child-safety reporting.

The Council's figures are AI simulations. The historical figures on the Council are written by an AI from public sources. They are not the real people. Neither the people nor their estates are connected with us, and none of them has endorsed the app.

What the companies may and may not do.

Babushka is an AI. She is a character, not a person. She is not a doctor or a therapist.

Where the conversation is kept. Your conversation with her, and the Council's answers, are stored on our servers (§8), not by the AI companies. That's how the thread is still there when you come back.

Health, relationships and other sensitive topics. People tell Babushka about their health, their relationships, their sex life, their feelings and their beliefs. The law calls this sensitive or special category data. When you say yes on her card, you give your explicit consent for her to read such things, and to remember them, only to talk with you. We use them for nothing else. Before she keeps a sensitive note about you, she asks you in the conversation (§5).

Her overnight thinking. Between conversations, Babushka may read your conversation and her notes again, to update what she knows and think about what to say next. This also runs through xAI, under the same rules, and only while her room is open for you.

Where the helpers aren't available.

We go by your App Store country and by the country your connection comes from: if either is on a list, that helper stays closed. Where a helper is closed, the app says so or does not offer it, nothing you type is sent to any AI company, and the rest of the village works.

If we change an AI company. We ask again only for a new AI company or a new kind of data. If we move a helper to a different company, or want to send it a new kind of data, we update this policy, and the app asks you again, on a card that says what changed, before anything goes to it. Your earlier yes covers only the company and the data it named.

5. Babushka's notes about you

6. Counting how the app is used, and crash reports

6.1 Usage counts

What we count. Which parts of the app you open, what you tap there and for how long; lessons, games and snacks started, finished, answered (right or wrong) and left; your answers on our permission cards; errors you meet; how long the app takes to start and to load, and whether a video stalls; how many messages you send to the AI helpers and how long their replies take (never what either side says); the app version; your device model, iOS version, language setting, time zone, screen size and network type (Wi-Fi or mobile); your App Store country; and whether you are an adult or a teenager (never your birth year). Counts are linked to a random app ID, not to your name. They never include what you write or say. From your phone's internet (IP) address, PostHog also estimates roughly where you are (the country, region, city and postcode area, never your exact position), and then discards the address instead of storing it.

Who counts. PostHog, an analytics service, in the EU (Germany). In Settings this is the switch "Help us improve the app". PostHog also receives anonymous daily totals from our own database, such as how many lessons were finished on a day: statistics with no account ID, in which any group of fewer than five people is left out. The second switch, "Let me get to know you", lets us keep the activity in §3.3 in our own database, to choose your next lessons, and a copy of your guest book (§3.2); it is always a yes-or-no choice.

How we ask. It depends on where you are and on your age.

What a no, or the switch turned off, leaves. The app still keeps your progress, because the game runs on it (§3.3), and our server still turns progress into anonymous totals with no account attached, such as "this lesson was finished 4,000 times" (§9). Nothing new is collected for counting.

What we record for the people we count. Which screens you open and how long you stay on each, linked to the random app ID. We use it mainly as totals, to see patterns like "most people stop at this step", and we may look at one person's own path through the app to find where it goes wrong for them. We never connect it to your name.

Screen recordings. The app never records your screen, in any version, test versions included.

6.2 Crash reports

If the app crashes or freezes, a technical report goes to Sentry, in the EU (Germany). It says which part of the code failed, and includes your device model, iOS version, app version, a random install ID, and the last few network addresses the app called (these can include your random account ID). It never includes a screenshot or anything you wrote, and in the App Store version no name or account is attached to it. Crash reports start the first time you open the app, before any card asks you anything, so we can see a crash on the way in. They are on unless you switch them off: Settings → Privacy & account → What Babushka may know → Crash reports (in a test version they stay on, §6.3). Once the app knows you are under 13, none are sent, in any version of the app, and the switch stays off (§13); before it knows anyone's age, a crash report can be sent, with no name attached.

6.3 Test versions of the app

If you test a pre-release version (TestFlight), we record more, so we can find problems: every screen you visit, how long you stay, where you leave the app, and how long parts of the app take to load, linked to your random account ID, and crash reports carry that ID too. This starts when you first open the test version, because each tester agrees in writing before installing it, and we tell testers in the invitation; our record of your permissions notes that written yes. You can switch it off in the test version too (Settings → Privacy & account → What Babushka may know), and then nothing more is counted. Crash reports stay on in a test version. A test version never records your screen either.

7. Who handles your data for us

These companies process data only on our instructions, under a data-processing contract. This table is the current list; if it changes, we change this page first.

Company What they do for us What they get Where
Supabase Our database, accounts and server code Everything in §3, §4 and §5 that we store Ireland (EU)
Cloudflare Delivers the app's pictures, sound and video; hosts our website and passes on the email you send us; relays your voice in the Theatre's Scenes to OpenAI Your IP address when your phone downloads a file or you open our website; the emails you send us, on their way to our mailbox; your voice in the Scenes, on its way to OpenAI, with no copy kept Worldwide network
PostHog Usage counts (§6.1) Counts linked to a random app ID; your IP address, which it uses to estimate your city and then discards; anonymous daily totals from our database, with no account IDs Germany (EU)
Sentry Crash reports (§6.2) Crash reports Germany (EU)
xAI Babushka's replies, her safety checks, your alias suggestions and the Theatre's notes (§4) See §4 USA
OpenAI The Council, and the other person in the Theatre's Scenes (§4) See §4 USA

Apple delivers the app and tells the app your App Store country. Apple's App Attest and DeviceCheck tell our server that the app is genuine and running on a real iPhone (§3.1). Apple is responsible for its own use of data; see Apple's privacy policy.

Public archives. If you play a historical recording in the Theatre's Great Speeches, your phone streams it straight from the public archive that holds it (for example the US National Archives or the John F. Kennedy Presidential Library), which sees your internet address as any website would. We send them nothing else.

Nobody else. We don't give your data to advertisers or data brokers.

Only where the law makes us. We never report you to the police or emergency services for anything you tell us about yourself, and we never contact your family. We disclose data only where a law that applies to us requires it:

We tell you when the law allows us to.

If the company is sold, your data could pass to the buyer, who would have to keep to this policy or tell you about any change first.

8. Where your data is, and when it leaves the UK and EU

9. How long we keep things

What How long
Your conversation with Babushka Each message: 365 days, then deleted automatically. Or until you delete the conversation.
Babushka's notes about you, and her summary of you Until you delete them, ask her to forget everything, or delete your account
Her copy of your Notebook notes Until you mark a note Private, delete it, turn off her reading, or delete your account
Your Notebook Until you delete it. A deleted note's scrambled text is erased 30 days later.
The copy of your guest book, and your alias Until you delete your account, or ask us to delete the copy
Reflections, Theatre takes, your Council Ledger, your progress Until you delete them or your account
Activity (§3.3) One year. Then, a whole month at a time, the records are turned into anonymous totals, such as "this lesson was finished 4,000 times", and deleted, so a record lives at most 13 months. When you delete your account, your answers are turned into anonymous totals at once, and the rows tied to you are deleted.
What the feed served you: the list of what it put in front of you, and the feed's own log of what it chose to show, which carries no account ID Two years, then deleted. The list also goes when you delete your account; the log cannot be tied back to you once your account is deleted.
A record of each call to an AI company: which feature made it, the model, how much text went in and out, and what it cost; never what you or Babushka said Two years, then deleted, or when you delete your account
Reports you send about her replies or a Council answer Until we close the report; then two years from the day it is closed
Safety records (§15) While your account exists. If our safety check stops one of her replies, the words of that reply are erased after 30 days.
A note of hers you corrected or tore out (§5) The old words stay out of her sight, as a guard against her learning them again, until you delete the note in the dossier, ask her to forget everything, or delete your account
An idea you forget in Your file Kept for 30 days so you can put it back, then erased; a marker with no content stays, so that older answers cannot teach it back
The security key for the app on your iPhone (§3.1), your age correction (§2), your time zone and when you last opened the app (§3.3) While your account exists
Your permission choices (which card you agreed to, which words it showed, and when) While your account exists. When the account is deleted, a minimal proof of each choice stays for three years: the kind of choice, the version of the words, when it was given and taken back, and a one-way scrambled form of the account ID, with nothing else about you. Then deleted.
A photo you show Babushka Deleted as soon as she has answered. A refused photo is deleted at once and never kept. One you chose and never sent is deleted within two hours. We keep no copy.
What a report to NCMEC contains (§4) One year from the report, because US law treats a report as a request to keep its contents. If you delete your account in that year, we keep only what the report contained (never a picture), locked away where no app or key can reach it, until the year is up. Then deleted.
A child-safety flag we have not yet decided on (§4) Until a person decides, if you delete your account first: the flag only (never a picture), locked away; then reported (kept as above) or deleted.
Your acceptance of the Terms (which versions of the Terms of Service and this policy you accepted, the app build, and when) While your account exists; then as a minimal proof for three years, like your permission choices
Copies at the AI companies xAI: about an hour. OpenAI: up to 30 days (the Council, and your voice in the Theatre's Scenes).
Usage counts (PostHog) As long as PostHog's plan keeps them: one year on its free plan, seven years on a paid plan. PostHog offers no shorter period.
Crash reports (Sentry) Up to 90 days, then Sentry deletes them
Our servers' technical logs, which include IP addresses 7 days
Backups Our database is backed up daily, and each backup is kept for 7 days. A deleted item can live on in a backup until that backup expires.
An account nobody opens Deleted, with everything in it, once it has not been opened for 12 months, exactly as if you had deleted it yourself
Emails you send us As long as we need them to answer you, and no longer than two years after your last message

If you delete the app without deleting your account, your data stays on our servers until the account has gone 12 months without being opened; then it is deleted. Because your account is anonymous, we can't find it again for you once the app is gone. So please delete your account in the app first (Settings → Privacy & account → Delete account).

10. What we never do

  1. We never sell your data, and never share it for advertising. There are no ads in the app.
  2. We never track you across other apps or websites. We don't use an advertising ID.
  3. We don't train AI models on your conversations. We don't collect, use or sell personal data to train large language models. If we ever want to, we will ask you separately. Only adults will be asked. Nothing will be ticked for you, and saying no will change nothing else. Nothing collected before your yes will be used.
  4. The AI companies may not train on your words either (§4).
  5. No one on our team reads your conversations as a matter of course. A person sees part of one only when you send us a report (and only the part you chose), when our safety check stops one of her replies (we see her reply, not your words), or when the law requires it. Our small team can reach the database to keep the service running, and that access is limited.
  6. We never report you to anyone for what you tell us, even when a safety check fires: not the police, not emergency services, not your family (§15). The only disclosures we make are the ones the law requires (§7); for photos, that means the reports about child abuse images in §4.
  7. We never guess your emotions from your face, voice or typing.

11. Your rights, and how to use them

Most of this you can do yourself, in the app, right now.

You can In the app By email
See everything we keep about you Settings → Privacy & account → Your file, or the folder on Babushka's table Yes
Get a copy of your data Settings → Privacy & account → Download my data Yes
See what Babushka knows about you Settings → Babushka → The dossier Yes
Correct something Correct a note in the dossier Yes
Delete one note, all her notes, or the conversation Settings → Babushka Yes
Delete your account and everything with it Settings → Privacy & account → Delete account Yes
Turn off an AI helper, usage counts, crash reports or personalised lessons Settings → Privacy & account → What Babushka may know Yes
Object to a use, or ask us to restrict it Yes
Take your data to another service Download my data gives you a file Yes

Deleting your account deletes everything we hold about you on our servers, and the copies on your phone, except, if we made one, what a report to NCMEC contains, locked away until a year after the report, or a child-safety flag still waiting for our decision (§4, §9), and the minimal proof of your permission choices, kept three years with no way back to you (§9). Copies at the AI companies expire as in §9. Usage counts and crash reports are not linked to your account, so we can't pick yours out. Nor is the feed's own log of what it chose to show, which carries no account ID and is deleted after two years (§9).

Writing to us. Email info@kgbbabushka.com. Because your account is anonymous, we may ask you to make the request from inside the app, or to send the account code shown at the bottom of Settings, so that we act on the right account. We won't ask for anything more than we need.

How fast. We answer within one month. If a request is complex, we may take up to two more months, and we'll tell you why. It's free.

Complaints. If you're unhappy, please tell us first. You can always complain to the UK Information Commissioner's Office (ico.org.uk), or to the data protection authority where you live or work in the EU.

What we do Our legal reason For sensitive data
Run your account, the game, your progress, your Notebook, reflections, takes and Council Ledger It's needed to give you the service you asked for (Art. 6(1)(b)) —
Keep whether you are an adult or a teenager The age rules are a legal duty (Art. 6(1)(c)), and it's needed to give you the service (Art. 6(1)(b)) —
Keep a copy of your guest book on our server Your consent, "Let me get to know you" (Art. 6(1)(a)). You can take it back at any time. —
Reserve your alias It's needed to give you the service you asked for (Art. 6(1)(b)) —
Send your words to an AI helper, and keep the conversation so it's there when you return Your consent on the card (Art. 6(1)(a)) Your explicit consent on the same card (Art. 9(2)(a))
Keep your time zone and when you last opened the app, for her daily thought Your consent on her card (Art. 6(1)(a)). You can take it back at any time. —
Everyday notes about you Our legitimate interest in making Babushka work as described (Art. 6(1)(f)). You can turn them off. —
Her copy of your Notebook Your consent: her card, and then "Let her" (Art. 6(1)(a)). You can take it back at any time. Your explicit consent, given by "Let her" after her card (Art. 9(2)(a))
Sensitive notes about you Your consent (Art. 6(1)(a)) Your explicit consent, given in the conversation (Art. 9(2)(a))
The safety check on every message (§15) Legal obligation: companion-chatbot laws require a crisis protocol (Art. 6(1)(c)); protecting someone's life (Art. 6(1)(d)) Your explicit consent on her card (Art. 9(2)(a))
Checking a photo for child abuse, and reporting a clear child to NCMEC (§4) Our legitimate interest in keeping child abuse images off our service and helping protect children (Art. 6(1)(f)) In the UK, for data about a possible crime: preventing or detecting unlawful acts, and safeguarding children (DPA 2018, Sch. 1 paras 10 and 18), under our appropriate policy document
Usage counts for adults whose App Store account is in the United States or the United Kingdom (§6.1) Our legitimate interest in understanding how the app is used and making it better (Art. 6(1)(f)), after a clear notice; you can object at any time with the switch. In the UK, storing and reading this on your phone for statistics needs no consent when there is a clear notice and a simple way to object (the Privacy and Electronic Communications Regulations, as amended by the Data (Use and Access) Act 2025, in force since 5 February 2026) —
Usage counts for everyone else, and lessons chosen from your behaviour Your consent, on the card or the switches (Art. 6(1)(a)). You can take it back at any time. —
Anonymous totals made from your progress, and from activity older than a year, including the daily totals sent to PostHog Statistics compatible with running the game (Art. 5(1)(b) and Art. 89); no account is attached —
Keep a minimal proof of your permission choices for three years after the account is deleted Our legitimate interest in being able to show that we asked, and in answering a complaint or a claim (Art. 6(1)(f); Art. 7(1)) —
Apple's age range, where Apple says the law requires it Our legitimate interest in applying the age rules, and the laws that require us to check (Art. 6(1)(f)); used on your phone and never stored; in an age correction only whether it says 18 or over reaches our server, and it is not kept —
Keeping a tester's written yes in our records of permissions Our legitimate interest in being able to prove that we asked (Art. 6(1)(f)) —
Crash reports Our legitimate interest in fixing faults (Art. 6(1)(f)). You can switch them off. —
Keep a record that you accepted the Terms (the versions, the app build and the time) Our legitimate interest in being able to show which terms you accepted (Art. 6(1)(f)) —
Keeping the service secure and stopping abuse, including App Attest and DeviceCheck (§3.1) Our legitimate interest (Art. 6(1)(f)) —
Keeping records the law requires Legal obligation (Art. 6(1)(c)) —

No decision with legal or similarly serious effects on you is made by automated means alone. Lesson suggestions and the safety checks are automated. One automated result has an effect: a photo that our check reads as clearly showing a child in sexual content, or a drawing it reads as clearly sexualising a child, closes photos for that account at once. Whether it is reported to NCMEC (§4) is decided by a person.

13. Children and teenagers

14. If you live in the United States

Everyone. We don't sell your personal information and we don't share it for targeted advertising, under any state's law. There is nothing to opt out of, but we honour a Global Privacy Control signal anyway. Whatever state you live in, you can ask to know what we hold, get a copy, correct it and delete it, and you won't be treated worse for asking. Use the in-app controls (§11) or email info@kgbbabushka.com. You can use an authorised agent; we'll ask for proof. If we refuse a request, you can appeal by emailing us with "Appeal" in the subject. We answer appeals within 45 days.

California. In the last 12 months we collected: identifiers (a random account ID, the security key for the app on your iPhone, and the name you wrote in the guest book, if it was copied to us); whether you are an adult or a teenager, and the year you were born if your guest book was copied (§3.2); internet and app activity (§3.3, §6); audio only if you turn on Save takes to my account (§3.3), or once the Theatre's voice rooms open (they are closed in version 1.0); inferences (Babushka's notes, lesson suggestions); and sensitive personal information: the contents of your messages to the AI helpers, which may reveal health or sex life. We collected it from you and from your use of the app, for the purposes in §3–§6, and disclosed it only to the companies in §7. We use sensitive personal information only to provide the service you asked for, so we don't offer a "limit the use" link. Our crisis protocol, which California law requires us to publish, is at kgbbabushka.com/crisis/; §15 is its summary.

Connecticut. We don't collect, use or sell personal data to train large language models. We process sensitive data only with your consent (§4, §5).

Washington and Nevada. Some of what Babushka holds, such as a note about your health, is "consumer health data" under their laws. Our Consumer Health Data Privacy Policy is at kgbbabushka.com/consumer-health-data/.

Other states with privacy laws (such as Virginia, Colorado, Texas, Oregon and others) give you the rights above. Where a state requires consent for sensitive data, the cards in §4 and §5 ask for it.

15. Safety, and what happens in a crisis

Every message you send Babushka is checked automatically for signs that you may be in danger. The check runs on xAI, the same company as her chat, under the same rules. Beside it runs a fixed set of rules on our own server, with no AI. The plainest statements of danger, such as a stated plan to end one's life, a plan to jump from a height today, or having taken ten or more pills, always get the crisis reply, whatever the check says. A wider set of rules catches the plainest words of danger when the check cannot run, or while her chat is closed or paused.

If the check or a rule fires, she answers with words a person wrote in advance, and the app shows a card that points to findahelpline.com, a free directory of helplines for nearly every country (its page for your country, where it has one), and, if your App Store account is in the United States, to 988, the Suicide & Crisis Lifeline (call or text). We print no other helpline numbers. If you tell her you have already taken an overdose, she first tells you to call your emergency number now (for example 999 in the UK, 911 in the United States and Canada, 112 in the EU) or to get someone to take you to an emergency department. The fixed rules and the card work even when the check, or the whole connection to xAI, is down or switched off.

We never contact the police, emergency services or your family about what you tell her about yourself. We show you where help is. The only disclosures we ever make are the legal duties in §7, which are about crimes, never about someone who is struggling.

We record that the check fired, the kind of concern, and when. This stays with your account until you delete it. From July 2027, California law requires us to report once a year how many times we pointed people to crisis services, as counts with no names and no content.

Our full crisis protocol, which California law requires us to publish (Business and Professions Code §22602), is at kgbbabushka.com/crisis/.

16. How we protect your data

If a breach puts you at high risk, we'll tell the ICO within 72 hours and tell you as soon as we can, in the app, in plain words.

Our website has no analytics, sets no cookies and loads nothing from other companies; even its font is served from our own site. Cloudflare hosts it (§7).

17. Changes to this policy

We put each new version on this page, with its date. Earlier versions stay readable: the April 2026 version is at kgbbabushka.com/privacy/2026-04/.

We don't have your email, so the app is where we tell you.

18. Contact

info@kgbbabushka.com · Pungent Ltd, 167-169 Great Portland Street, London, W1W 5PF, United Kingdom · EU representative: being appointed (§1) · ICO registration: registration in progress (application ref C2053707) · Person in charge of the protection of personal information (Quebec): the Director, Pungent Ltd, at the same addresses

Version 2.0 replaces the April 2026 policy. It adds the AI helpers, the guest book, usage counts, crash reports, anonymous accounts and the retention table, and drops email accounts and purchases.