Skip to content

Privacy

Privacy policy

Last updated: 16 November 2026

You talk to Hanaso out loud and your coach remembers you, so we owe you a clear account of what we keep. This page explains what we collect, why, who processes it for us, where it goes, and how to see, export or delete it.

1. Who we are

Bypass Inc. ("Hanaso", "we") operates the Hanaso apps for iPhone, iPad, Apple Watch and Android, the Hanaso web app, the messenger bots and the website hanaso.io. This policy covers all of them.

It is written to meet Japan's Act on the Protection of Personal Information (APPI), the EU General Data Protection Regulation and the UK GDPR (together, "GDPR"), the California Consumer Privacy Act (CCPA) and the rules of the App Store and Google Play. We publish it in several languages; if a translation differs from the Japanese version, the Japanese version prevails.

2. What we collect

We collect what the features you use need, and we ask for permissions at the moment they're needed.

  • Account: email address and a password stored only as a secure hash, or the identity token from Sign in with Apple or Google.
  • Profile: display name, handle, the language you speak and any others, country, year of birth (to apply age rules), time zone and interface language.
  • Learning: target languages, goals and dates, level, Speak Rating, flashcard reviews, scenario results, mock exam answers and scores, pronunciation scores, corrections and progress.
  • Conversations: what you type to your coach, transcripts of what you say, and your coach's replies.
  • Voice: audio of your speech, processed to understand you and to score pronunciation (see section 3).
  • Your coach's memory: facts and short summaries drawn from your conversations (see section 4).
  • Social: friends, squads, duels, help board posts and answers, club posts, cheers, gifts, reports and blocks.
  • Purchases: your plan, purchase and subscription status from Apple, Google or Stripe through RevenueCat. We never see your full card number.
  • Device and notifications: device type, app version, push token and notification settings.
  • Messengers: if you link Telegram, LINE or WhatsApp, the identifier that service gives us and the messages you send to the bot.
  • Waitlist: email address, the languages you chose, your goal, an optional exam, the page you joined from and an invite code if you had one.

3. Your voice

When you talk, your audio is streamed to our servers and to speech partners to turn it into text, generate the coach's reply and score your pronunciation.

  • Recordings are stored only if you turn on "Keep voice recordings", for example for "then vs now" clips. You can turn this off and delete all recordings in Settings under Privacy and data.
  • Your voice is never shared with other learners unless you record a duel answer or turn on "My voice" in a clip you choose to share.
  • Talk table and tandem transcripts are kept for 30 days for safety reports only, then deleted.
  • A child's voice audio is never stored (see section 11).

4. Your coach's memory

To make conversations personal, your coach keeps facts about you (for example your job, people in your life, goals and upcoming events), short summaries of your days and a profile card describing how you like to learn. They are stored with mathematical representations (embeddings) so the coach can recall them.

You can see every fact in "What your coach remembers", edit or delete any of them, delete everything, or switch memory off. Switching it off during your first conversation also clears what was extracted. Memory is never shown to other learners.

5. How we use your information

  • To run conversations, role-plays, flashcards, pronunciation, exams and your daily plan.
  • To remember you, if memory is on.
  • To run the game and social features: XP, petals, Speak Rating, leagues, squads, duels and matchmaking by level, age band and time zone.
  • To keep people safe: moderating what people write or say to each other, handling reports and enforcing our rules.
  • To process purchases and apply your plan's limits.
  • To send notifications you allow, including the coach call, and emails about your account.
  • To improve the quality of coaching, using learning data (for example which corrections help), never selling it.
  • To measure which of our ads bring learners to the iOS app (see section 9).
  • To meet legal obligations.

Where GDPR applies, our legal bases are contract (running the service you signed up for), consent (voice recordings, memory, transfers to AI partners, marketing email), legitimate interests (safety, fraud prevention, product quality) and legal obligation.

For ad measurement in the iOS app, the legal basis is your consent when you allow tracking, and our legitimate interest in measuring our ads when you don't (then without the advertising identifier).

6. AI processing

Hanaso's coach runs on large language models from Anthropic and OpenAI. OpenAI also provides speech-to-text, text-to-speech and real-time voice conversation, and the embeddings your coach uses to recall memories. Pronunciation is scored on our own servers or by a pronunciation-assessment provider. What you say or type, the context your coach needs (including relevant memories) and your coach's replies are sent to these partners to generate responses.

Under their API terms, Anthropic and OpenAI do not use data sent through their APIs to train their models by default. We do not use your conversations to train any model that identifies you.

7. Transfers outside your country

Our database is in Tokyo, Japan. Some partners process data in other countries, including the United States. Before any of your words reach an AI partner, we ask for your consent in plain words, as Japanese law requires. Where GDPR applies, we rely on adequacy decisions or standard contractual clauses for these transfers.

You can withdraw this consent, but Hanaso's conversation features cannot work without it.

8. Who processes data for us

  • Supabase (database, Tokyo region) and Cloudflare (hosting, storage, delivery and email).
  • Anthropic (language models) and OpenAI (language models and embeddings).
  • OpenAI (speech-to-text, text-to-speech and real-time voice in every language we support). Pronunciation is scored on our own servers or by a pronunciation-assessment provider.
  • RevenueCat, Apple, Google and Stripe (purchases and subscriptions); Wise (creator payouts).
  • Apple and Google (Sign in with Apple and Google sign-in, if you use them).
  • Google Firebase (push notifications and app analytics).
  • Meta and TikTok (measuring our ads in the iOS app, see section 9).
  • Telegram, LINE and WhatsApp, only if you link a messenger.

We do not sell your personal information. Apart from ad measurement in the iOS app, where you decide whether to allow tracking (see section 9), we do not share it for cross-context behavioural advertising.

9. Analytics and cookies

Our apps and web app use Google Analytics through the Firebase SDK to understand how features are used, for example which onboarding step people leave at. Kid profiles send no analytics.

The iOS app also uses software from Meta (Meta Platforms, Inc.) and TikTok (TikTok Pte. Ltd.) to measure our ads: which ad brought you to Hanaso, then app opens, finishing onboarding and your first talk, level-ups, streak milestones, badges, views of the plans screen and purchases you start or make in the App Store (product, price and currency). No name, email, user ID or conversation is sent. Once, about a second after the lobby first opens, the app asks signed-in learners aged 13 or older whether to allow tracking. If you allow it, Meta and TikTok can use your device's advertising identifier to match these events to their ads; if you don't, both run in a limited mode without it. Nothing is sent from kid profiles or for learners under 13. You can change your choice at any time in iOS Settings under Privacy & Security, then Tracking.

This website uses Cloudflare Web Analytics, which does not use cookies or track you across sites. The website sets a cookie to remember your language choice; the web app sets cookies to keep you signed in.

10. What other learners see

Friends see your display name, coach, languages, streak, league tier and pinned badges. Your Speak Rating, goal, garden and moments are shown only if you switch them on. Transcripts, audio and your coach's memory are never shown.

Social features are available from age 13. Learners aged 13 to 17 meet only other learners aged 13 to 17 in live features and with strangers.

11. Children

Children aged 6 to 12 use Hanaso only in kids mode, inside a parent's account. A parent gives verifiable consent before any child's voice is processed, using a method chosen for their region.

In kids mode: a child's voice audio is never stored; memory keeps learning facts only (level, words, mistakes and interests the parent allows), never personal-life facts; social features, sharing, messengers and purchases are closed; and no analytics are sent. Parents can view and delete what the coach remembers, and remove the child's profile, in the parent dashboard.

We follow COPPA in the United States, GDPR rules for children, APPI, and App Store and Google Play rules for apps used by children. If you believe a child is using Hanaso outside kids mode, contact us.

12. How long we keep data

  • Account, profile and learning data: while your account exists.
  • Conversation messages: 18 months in our main database, then archived or deleted.
  • Voice recordings: only while "Keep voice recordings" is on, and until you delete them.
  • Talk table and tandem transcripts: 30 days, for safety reports.
  • Purchase records: as long as tax and accounting law requires.
  • Waitlist entries: until 12 months after launch, or until you ask us to remove you.

When you delete your account, we delete or anonymise your data as described on the "Delete your account" page.

13. Your rights and choices

  • See and download your data with "Export my data" in Settings.
  • Correct your profile, and edit or delete memories, in the app.
  • Delete voice recordings, memories or your whole account.
  • Withdraw consents in Settings under Privacy and data.
  • Object to or restrict some processing, and complain to your data protection authority.
  • California residents can ask what we collect and ask us to delete it. We do not sell personal information, and apart from ad measurement in the iOS app, where you decide whether to allow tracking, we do not share it.

For anything you can't do in the app, email [email protected]. We may ask you to confirm your identity.

14. Security

Data is encrypted in transit and at rest. Passwords are stored only as salted hashes. Learner audio is kept in private storage and opened only through short-lived signed links. Access by our staff is limited by role and logged.

15. Changes to this policy

If we make material changes, we will tell you in the app or by email before they take effect, and ask for new consent where the law requires it.

16. Contact

Privacy questions and requests: [email protected]. Data controller: Bypass Inc., Japan. Representative and personal information manager: Yousef Kenji (剣地 ユセフ). Our address is disclosed without delay by email on request. Residents of the EU and the UK can also contact us at [email protected].