Privacy Policy
Last updated November 4, 2026
Takaki Takemura, an individual developer in Japan ("we"), makes Crayon Study: Language ("the app"), a language-learning app for iPhone and Android in which a crayon-drawn buddy and AI tutors help you learn Japanese, Korean, Chinese, English, Spanish, French, German and Italian. This policy explains what information the app handles, and why.
What stays on your phone
Your goals and test dates, your study records, the lessons you have done, the words and sentences you keep, your chats, your buddy and its outfits and furniture, and your settings are kept on your phone (and in your phone's own backup). They are not sent to us. There is no account and nothing to sign up for: the app never asks for your name, email address or phone number.
The app's random ID
The first time you use the AI, the app makes a random ID for your phone (never for a user Apple's age range says is under 13, below). It contains nothing about you. The app keeps it on the phone, and on iPhone also in the iOS Keychain, so that the chat tickets you bought are still there if you reinstall the app on the same iPhone. It is sent with each request to our server, which keeps only a one-way hash of it, to count each day's questions and to keep your chat tickets and the coach's answers.
When you use the AI
Looking things up, chatting with your buddy and the friends, "Say what you mean", the teacher, lessons, dramas, this week's words and the AI coach are answered by our server (Google Cloud, Tokyo) with Claude, an AI by Anthropic. For each request, the app sends:
- what you typed or chose to look up (also text read from a photo or the camera on your phone), your message and the recent lines of the chat, or what the coach is asked about (your study time and results of the week, the plans you entered, and in the speaking test the text of what you said);
- your buddy's name and whether it is a dog, a cat or a plant, the level and scene you chose;
- the language you are learning, the language of the app, the way of speaking you chose, whether dirty mode is on, and your age range (below);
- the random ID above, and, if you have a plan, the store's proof of purchase (below).
The app never sends your voice itself (only the text your phone made of it), and never your name, email address, date of birth or location. The random ID goes only as far as our server: it is not sent to Anthropic.
Our server passes what is needed to Claude, by Anthropic, PBC, to write the answer, and the text to be read aloud to Google's Text-to-Speech (Gemini-TTS) to make the AI voice. Our server does not keep what you asked or what was said in your chats, and does not write it in its logs, except as listed below. Anthropic handles what it receives under its commercial terms, does not use it to train its models, and deletes it after a limited period, under its own policies.
Lessons, dramas, this week's words and the coach's practice sets are written by the AI once and shared by everyone. They are not about anyone and hold nothing of yours.
What our server keeps, and for how long
| What | Linked to | Kept for |
|---|---|---|
| How many questions, lessons, voices, dramas, coach answers, ads for an extra question and reports your phone used today | The hash of the random ID | 3 days |
| The coach's answers to you: the week's review, the day's check and speaking test, and which test topics you have had | The hash of the random ID, and the language | 14 days (the week's review), 2 days (the day's check and test), 400 days (the list of topics) |
| The chat tickets your phone bought and has left, and each ticket purchase (the store's transaction number, the product, the number of questions, the date) | The hash of the random ID | As long as the service runs, as a record of what was paid for |
| The answer to a look-up made without surrounding text, so that the same look-up by anyone is answered at once (it contains the words looked up) | Nobody | 30 days |
| Voices made by the AI (MP3) | Nobody | 30 days; a lesson's lines as long as the lesson |
| What you report (below) | Nobody (not the ID, not the age range) | 180 days |
| Everyone's totals of the day, overall and by language | Nobody | 90 days |
| What the AI and the voices cost each day, by language (counts, tokens, seconds, amounts) | Nobody | 400 days |
These are deleted automatically when their time is up.
Reports
If an answer seems wrong or inappropriate, you can report it from the flag beside it. What you send is kept: the part of the app, the answer, the text it was about, the reason you chose and any note you write. We read it to correct the AI's instructions and to remove what should not be there. It is not linked to your phone or your age, and is deleted after 180 days. A phone can send up to 20 reports a day.
Your age
The app is rated 13+. To keep the AI's answers right for younger learners and to keep dirty mode for adults, the app may know your age range, never your birth date:
- On iPhone with iOS 26 or later, the app asks iOS once, through Apple's Declared Age Range, which only tells the app whether you are under 13, 13 to 15, 16 to 17, or 18 or over, and only if you agree to share it.
- When you turn on dirty mode where iOS has not told the app, you are asked whether you are 18 or over. This is the only time the app asks you.
Under 13. If Apple's age range says you are under 13, the app does not call our server at all and sends no ID: it shows a kind screen saying the app is for ages 13 and up, with a line for parents. Nothing about you reaches us.
The age range is kept on your phone and sent with each request to our server, which uses it to choose the AI's rules (the rules for under 18 also apply when the age is unknown) and to refuse dirty mode to anyone not known to be an adult. The server does not keep it; its logs may note the range, never together with anything you wrote.
Voice, microphone, camera and photos
- To check your pronunciation and to let you speak instead of typing, the app uses the microphone and your phone's own speech recognition (Apple's on iPhone, the phone's speech service on Android). Depending on the phone, the sound may be recognized on the phone or by Apple or Google, under their policies. We do not receive your voice. The pitch line is drawn on the phone.
- To look up text in a photo or through the camera, the text is read on your phone (Apple's Vision on iPhone). The picture is not sent anywhere; only the words you then choose to look up are sent, as above.
- Reading aloud in the phone's own voice happens on the phone. The AI voice is made by our server, as above.
Crayon Pets and Crayon Plants
On iPhone, if you also have our apps Crayon Pets or Crayon Plants, the app can read how your pets or plants look from them, through storage the apps share on the phone, so that one can study with you. Nothing of this is sent to us.
Purchases
- Plans and chat tickets are sold by Apple (App Store) or Google (Google Play), who handle the payment. We never receive your payment details.
- To give you your plan, the app sends the store's proof of purchase (on iPhone, the purchase signed by the App Store; on Android, the purchase token) with your requests. Our server checks it with Apple's certificates or asks Google Play, keeps the answer in memory for a few minutes, and does not store it.
- For chat tickets, our server keeps the record in the table above, so that each purchase is credited once.
Ads
On the free plan, the app shows ads by Google AdMob: a banner on the home screen, a full-screen ad when you leave a finished lesson, and a video you can choose to watch for one more question. Ads are limited to content rated for all ages (G) for everyone. For users Apple's age range says are 13 to 17, ads are not personalized, and the app tells AdMob that the user is under the age of consent (its TFUA flag). Where the law requires it (such as in the European Economic Area and the UK), Google's consent message asks for your choice first. On iPhone, the app does not ask for permission to track you (App Tracking Transparency), so the advertising identifier is not used. On Android, Google may use the advertising ID as the phone's ad settings allow. AdMob may use information such as the IP address and the kind of device to show and measure ads and to prevent fraud. The paid plans have no ads.
Usage statistics and crash reports
To improve the app, it sends anonymous usage counts to TelemetryDeck (which screens and features are used, the language being learned, and, for a purchase, the product, its price and the country), without names or anything you wrote. It sends crash and error reports to Sentry (the device model, the system and app versions and the error), set up not to send personal information.
Notifications
Reminders are scheduled on your phone. We do not send push notifications and do not receive a notification token.
Server records
Like any web service, our server and Google Cloud record the requests the app makes (such as the time, the IP address and the kind of device) to run the service, find faults and prevent abuse. Our server's own logs hold what kind of request it was, the languages, the level, the plan, the age range, the amount of AI used and its cost, never what you asked. These records are kept for a limited time (30 days in Google Cloud's usual setting) and used for nothing else.
Third-party services
- Anthropic (Claude) — the AI's answers
- Google Cloud and Firebase — our server (Cloud Functions in Tokyo) and its database (Cloud Firestore)
- Google Cloud Text-to-Speech (Gemini-TTS) — the AI voices
- Google AdMob — ads on the free plan
- Apple App Store and Google Play — purchases
- TelemetryDeck — anonymous usage counts
- Sentry — crash reports
Each runs under its own privacy policy.
Providers outside Japan
We are in Japan. Some of the services above are run from other countries and handle what they receive under their own privacy policies and the laws of those countries: Anthropic, Google and Sentry in the United States, and TelemetryDeck in Germany (European Union). Our server itself runs in Tokyo. For the personal-information protection systems of other countries, see the materials of Japan's Personal Information Protection Commission.
The AI
- Your buddy, the friends, the teacher and the coach are AI, not people, and the app says so. They can make mistakes; check what matters in a dictionary or with a teacher.
- Please don't write personal details such as your full name, school, address or phone number in the chat.
- How the AI is kept safe, and what dirty mode is, are explained on Safety and AI.
Children and teenagers
The app is rated 13+ and is not directed to children under 13. As the US Children's Online Privacy Protection Act (COPPA) and similar laws require, we do not knowingly collect personal information from children under 13; the app asks for no name or contact details, and when Apple's age range says a user is under 13, the app makes no request to our server and sends no ID (above). For users 13 to 17, ads are not personalized. For learners under 18, or whose age the app does not know, the AI follows extra rules: it does not ask for personal information, does not take part in romantic role-play, and points to a trusted adult when someone seems to be in trouble. Dirty mode is for adults only. If you believe a child has sent us personal information, please write to us and we will delete it.
What we don't do
- We don't sell your information.
- There is no chat with other users, and no profile anyone else can see.
- The app does not use your location.
Your requests
You can ask us to tell you about, correct or delete what we keep about you, or to stop using it (rights given by Japan's Act on the Protection of Personal Information, the GDPR and similar laws), by writing to the address below. You can also delete it yourself in the app (Menu › Settings › Delete my data): see Delete your data. Most of it is deleted automatically within days (the table above). As we keep no name or email address, we may ask for details that help us find what is yours, such as the date and the store's order number of a ticket purchase. If you are not satisfied with our answer, you can also complain to the data protection authority where you live.
Changes
Changes to this policy are posted on this page, with the date.
Contact
Takaki Takemura, Japan