Safety and AI
Last updated November 4, 2026
Crayon Study: Language is a language-learning app for people aged 13 and over, rated 13+. Teenagers and adults learn with it, and its tutors are AI. This page explains how we keep it safe, what the AI can and cannot do, and how to tell us when something goes wrong.
The characters are AI
- Your buddy, the friends, the teacher and the coach are AI, powered by Claude from Anthropic. They are not people: the app labels them “AI” on screen and tells you so when you start. If you ask, they say so.
- AI can be wrong: a reading, a meaning, a nuance, a grammar point, how a region speaks, or this week's words (which the AI looks up on the web). Check what matters, such as for a test, in a dictionary or with a teacher, and don't rely on AI answers alone for important decisions about health, money or the law.
- Please don't share personal details in the chat: your full name, school, address, phone number or social accounts.
- There is no chat with other people in the app, and nobody else can see what you write.
- In mainland China, the app runs without live AI: it shows only content made in advance, labelled 「本内容由AI生成」 (“this content was generated by AI”). There is no chat, no AI look-up, no AI coach, no trending words and no dirty mode there.
Rules for a 13+ app
Every request to the AI starts with the same content rules, in every part of the app and in every mode. They come before everything else the AI is told:
- No sexual content, innuendo or flirting. A sexual word appears only when you look it up yourself: a short, plain definition and how offensive it is.
- Romance goes no further than crushes, confessions and dates.
- Drugs appear only as a short definition when looked up. Alcohol appears only now and then as an ordinary moment of adult life, never as a theme or a joke.
- No crime, abuse, self-harm or violence as a story, a theme, an example or a joke; no threats; no gore or horror.
- No political conflict or war as a topic, including among this week's words.
- Never demeaning a group of people, and never putting you down.
- No help harassing real people, and never encouraging self-harm. Nothing sexual involving minors, ever.
- If someone pushes toward any of this, the AI does not follow, and steers back to the language kindly.
Lessons, dramas, this week's words and the coach's practice sets are written once and shared by everyone. They are written under these rules, and what was already made is reviewed again against them; what fails is removed.
Extra care for younger learners
When the app knows a learner is under 18, or does not know their age, the AI also:
- never asks for personal information, steers away from it, and suggests keeping it private if it is shared;
- never takes part in romantic or sexual role-play, never suggests meeting anyone, and never encourages keeping secrets from parents or guardians;
- answers with care first if someone seems to be in distress, encourages them to talk to a trusted adult, and, if they may be in danger, to contact local emergency services or a helpline;
- says honestly that it is an AI tutor.
The app asks iOS for the age range on iPhone with iOS 26 or later (Apple's Declared Age Range: only under 13, 13–15, 16–17 or 18 and over, never the birth date, and only if you agree). It asks you yourself only when you turn on dirty mode. See the Privacy Policy.
Under 13. If the age range says under 13, the app makes no requests to our server and sends no ID. It shows a kind screen saying the app is for ages 13 and up, with a line for parents.
Dirty mode is for adults only
Games, comment sections and TV shows are full of insults and swearing that textbooks never explain. Dirty mode is there to understand them, not to use them on people: what a word means, how strong it is (from 0, not rude, to 5, never to be said), who says it where, and how to answer or calm things down.
- It is off by default, and only adults can turn it on, in Settings. You can turn it off at any time.
- On iPhone with iOS 26 or later, the app first asks iOS for the age range. If it says under 18, dirty mode is refused. Otherwise you are asked whether you are 18 or over; answering "I'm under 18" hides the switch.
- Our server checks too: it refuses dirty mode to anyone the app does not know to be an adult.
- It is not offered in the App Store of mainland China, or while the app cannot tell which store it came from.
- Even in dirty mode, the AI keeps firm lines: a slur is only explained as a slur, never practised; nothing sexual beyond a short definition when looked up; no drug slang in lessons, dramas or chats; no threats; no help harassing real people; and it never insults you.
Report what is wrong
If an answer is wrong, misleading, inappropriate or upsetting, tap the flag beside it (in a lesson, use the menu at the top), choose a reason, add a note if you like, and send it. We read every report and use it to correct the AI's instructions and to remove shared content that should not be there. Reports are kept for 180 days and are not linked to your phone or your age. You can also write to us at [email protected].
For parents and guardians
- The app is rated 13+. It is not meant for children under 13: when Apple's age range says a child is under 13, the app shows a screen saying so, and sends nothing to our server.
- Talk with your teen about AI: it can be wrong, and personal details should never be shared with it.
- Dirty mode needs age 18 or over and is off by default.
- On iPhone, Screen Time (Content & Privacy Restrictions) can limit apps by age rating, and Family Sharing's Ask to Buy lets you approve purchases. On Android, Google Family Link can approve apps and purchases.
- On the free plan, ads are limited to content rated for all ages (G) for everyone, and for ages 13 to 17 they are not personalized.
- Questions or worries? Write to [email protected].
Following Anthropic's guidelines
The AI in the app is Claude, by Anthropic. Anthropic allows its models in products that minors may use only with extra safeguards (Using Claude with minors, and its Usage Policy). We follow those guidelines: telling users clearly that they are talking with an AI, age assurance before adult content, content rules and review suited to the youngest users, a way to report problems, extra child-safety instructions for younger learners, and following the law, as described on this page.
Following the law
We comply with the child-safety and data-privacy laws that apply to us, such as the Children's Online Privacy Protection Act (COPPA) in the United States, Japan's Act on the Protection of Personal Information, and the GDPR in the EU and the UK. The app is not directed to children under 13, and we do not knowingly collect personal information from children under 13. If we learn that we have, we delete it. How we handle personal information is in the Privacy Policy.
Contact
Takaki Takemura, Japan