This policy explains what data ThaiHabit handles, why it is used and which services receive it.
Information we collect
The app stores learning progress, streaks, review history and preferences locally on your device.
If you start an online pronunciation exercise and separately allow Speech and pronunciation coaching, the recording and practiced Thai reference text are sent to our speech backend and Microsoft Azure Speech to transcribe or provide automated practice estimates. These estimates are coaching signals, not correctness judgments. Hugging Face ASR may be used server-side as a transcription fallback or optional comparison. These exercises are optional and require an internet connection.
The app may create a Supabase anonymous user id to enforce daily AI usage limits, restore Premium AI access and keep server-side quota records.
If you choose email, Google, or Sign in with Apple, Supabase stores the account email and linked provider identity needed to authenticate you and restore your account. Apple may provide a private relay email address.
Your selected Thai level, focus goal, polite ending and optional preferred name may be synced to Supabase so online conversations and phrase tools can use those choices across app sessions.
If you request new audio and separately allow Online reference audio, the exact Thai text is sent through a Supabase Edge Function to ElevenLabs. Audio generated from your added, imported or AI-created text is cached privately for your account, not shared with other learners. Approved recordings from ThaiHabit’s general learning catalog may be shared. The app does not use device speech as a substitute for Thai recordings.
If you choose a screenshot or take a photo for word import and separately allow Vocabulary image import, that image and text extracted from it may be processed by the app backend and Anthropic (Claude) or Hugging Face to extract and structure Thai words. This only happens when you start that import.
When you explicitly paste clipboard text or select a TXT, CSV, JSON, Anki, screenshot, or photo import, the app reads that selected content. Imported text is sent to Anthropic (Claude) or Hugging Face only after you separately allow Vocabulary text import and request an online cleanup or check; standard text and Anki parsing can stay on device. ThaiHabit disables Hugging Face Dataset row search that would place learner text in URL query parameters.
After you separately allow the applicable AI Conversation or Phrase Builder purpose, conversation messages or phrase instructions, custom goals, recent context and selected learning-profile context are sent through Supabase to Anthropic (Claude) to create the reply. OpenAI receives search text from that request to find matching reviewed Thai notes. Allowing one purpose does not unlock the other. Synced conversation threads and bounded response-cache entries may be stored in Supabase.
If you separately allow Online Thai language analysis, a typed Thai answer or lookup text and, for answer comparison, the expected answer and accepted alternatives may be sent through our backend to Hugging Face. Automatic wrong-answer feedback and live lookup remain local when this separate purpose has not been allowed.
After you separately allow Live talk and start a call, our backend sends your spoken turns to ElevenLabs for transcription. The transcript, recent call messages, chosen practice and relevant learning preferences go from our backend to Anthropic (Claude Sonnet) to create Ling’s reply.
When a call starts, the app also sends our backend a short summary of your learning status: your name in the app, how many items in each kind of practice are due for review and up to three Thai items you recently got wrong, such as a word or a letter. It goes to Anthropic (Claude Sonnet) with each request for a reply so Ling can address you by name and suggest fitting practice.
The summary can also include up to six Thai words or phrases you learned or saved in the last seven days and, if you opened Live talk from a practice, up to six Thai items from that practice, so Ling can bring them into the conversation. ElevenLabs turns the reply into audio. Ling is an AI tutor.
Your answers to Ling’s exercises on the screen, such as typed Thai, a sentence you put in order or an answer you pick, also go from our backend to Anthropic (Claude Sonnet) so Ling can respond to them. When Ling checks how you say a phrase, our backend sends that recording and the phrase to Microsoft Azure Speech, which estimates your pronunciation automatically. A short summary of that estimate goes to Anthropic (Claude Sonnet) so Ling can respond to it. This estimate is not a grade and says nothing about Thai tones.
Our backend does not store uploaded recordings. For call recovery, transcripts, exercise answers, pronunciation estimates, replies and reply audio are kept for up to 24 hours. After 90 days, a finished call keeps no speech, conversation or practice content and no learning status.
Only technical and billing records remain, such as call and turn IDs and times, how each turn ended, the call length and minutes used, plus the provider, model, processing time, amount of audio or text processed and cost of each processing step. Your purchase and minute balance records also remain. Providers may retain processed data under their own policies; our recovery limits do not set their retention periods.
When you use Live talk on iPhone, we temporarily store a technical token from Apple for the Live Activity on your Lock Screen. We use it only to end that Live Activity through Apple Push Notification service after your conversation is over and delete it once this is done. You can withdraw Live talk permission separately in Settings under Data and privacy.
Deleting your account removes your Live minutes, purchases and call records from ThaiHabit; it does not cancel a Store subscription.
If you contact support, confirm an error report, or send feedback in the app, the request is linked to your current app or account identity. Your message, optional email address, app version, platform and basic diagnostics are stored in Supabase and delivered to support through Resend so we can respond and fix issues. If you report an AI reply, the report also includes that reply and your message before it.
When you earn server-verified league XP online, your account (including a guest account) joins a weekly league group of up to 30 learners. When few learners are active, your group can include learners from nearby leagues. Others in your group see your league name, a random public id, your current league, your best league and your XP and place this week. Until you confirm a name, others see “Learner”. Your suggested name is the first name you entered during setup.
We store your league membership and weekly results, server-verified XP, name reports you make, learners you hide and league notifications we queue for you. If you win a Premium reward on a guest account, we keep it reserved under that account and add it only after you sign in. For answer verification we also store your account and request identifiers, the catalog and task identifiers, question and answer choices, your selected answer, the result and awarded XP, and issue, expiry and answer times.
During the switch to verified scoring we archive your previous open-week entry, group context and daily scoring records under your account. Other learners never see your email, sign-in id, answers, private progress or who reported a name. If three different learners report a name, it is hidden until its owner picks a new one or we restore it after review. We receive every report by email. In the league settings you can hide a learner, turn league notifications off or leave the league at any time.
We use the time zone of your most recently registered device to send league notifications during the day. We run the weekly league to provide the service you use (Art. 6(1)(b) GDPR). XP limits, name reports and hiding help keep it fair and safe, which is our legitimate interest (Art. 6(1)(f) GDPR). You can object at any time by leaving the league or contacting us.
If you set up Study buddies, we store your nickname, your invite code, your connections, whether you first joined through someone’s code, how many learners joined with your code and whether you share your profile. While you share, we also store your current streak and your latest learning day. Connected buddies see your nickname, streak and latest learning day.
If you pause sharing, the stored streak and learning day are deleted. If you remove a buddy, we keep a record of that so they cannot reconnect with your code. Reports about a buddy are sent to us for review.
If you turn on reminders or league notifications, we store a push token, a device ID, your language, time zone, platform and app version in Supabase so we can deliver them at a suitable time. Records for devices with notifications turned off are deleted 90 days after their last registration.
If you suggest a correction to learning content, we store the item it concerns, the text you saw and your suggestion with your account so we can review it. We delete them after 12 months.
Where handwriting recognition is available, Google ML Kit recognizes your strokes on your device. They are not sent to us. Google may receive technical requests when the recognition model is downloaded.
Our backend and hosting providers process network information for security, abuse prevention and rate limiting. Supabase stores hashed network subjects for bounded rate-limit records. When Sentry is enabled, it receives crash and app/device diagnostics with default personal data, tracing, profiles, request data and user identity disabled in app code.
The app stores your learning progress, including:
- Characters you have studied
- Review history and performance
- Learning streaks and statistics
- App preferences and settings
How we use your data
We use local and optional online data solely to:
- Track your learning progress
- Use your selected level and goals to suggest relevant Thai activities
- Remember your preferences and settings
- Assess pronunciation when you choose to record
- Provide cached AI reference audio and daily quotas
- Validate Premium access and restore purchases
- Respond to support or feedback requests you submit
- Run the weekly league, keep it fair and handle name reports
- Improve your learning experience
Data storage and security
Core learning progress, streaks, review history and preferences are stored on your device using the device’s built-in storage mechanisms.
If you sign in with email, Google, or Apple, eligible learning progress, preferences, custom word sets, their words and durable custom-vocabulary operation receipts are encrypted in transit and can synchronize to a Supabase account snapshot. Device-bound credentials, purchase caches, analytics identifiers, temporary import drafts and temporary media caches are excluded. Without account sign-in, core progress and custom vocabulary remain local-first.
You can reset local learning progress, delete local app data, leave the weekly league, export linked-account data or permanently delete a linked account and app-controlled cloud data from Profile. Leaving the league hides you from this week’s group and resets your XP for that week to 0. We keep your seat for the rest of the week so the results can be calculated. We store the final result with your other weekly results.
If you stay out until the week ends, you count as last place and move down one league if your group is ranked and your own league is above Bronze. Deleting your account deletes all your league data. If you use a guest account, deleting local app data also deletes its server data.
Third-party services
We do not show ads or use third-party behavioral advertising SDKs. On iOS, RevenueCat may collect Apple Search Ads attribution data to measure installs and subscription events. When AppsFlyer attribution is enabled, AppsFlyer may receive install, deep-link and privacy-filtered subscription-funnel event data. When product analytics is enabled, PostHog may receive privacy-filtered funnel events and allowlisted campaign labels, but not raw audio, transcripts, names, emails, support messages, raw URLs, or raw click ids.
The app uses
- RevenueCat for purchase-status verification and Premium access (it receives your app user id);
- Supabase for identity, account snapshots, support, conversation sync, quotas, the weekly league and Edge Functions;
- Cloudflare R2 for images, videos and audio storage and delivery, including generated reference audio;
- Resend for support notification emails;
- Anthropic (Claude) for separately permitted imports, phrases, conversations and word picture suggestions;
- OpenAI for searching reviewed Thai notes in separately permitted phrases and conversations;
- Microsoft Azure Speech for separately permitted transcription and pronunciation feedback;
- Hugging Face for separately permitted Thai text understanding, image reading, answer comparison and speech recognition;
- Google ML Kit for on-device handwriting recognition;
- ElevenLabs for separately permitted generated Thai audio where that feature is enabled;
- Expo, Apple, Google, Mozilla or another browser-selected push service for opted-in reminders and league notifications;
- Sentry for restricted crash diagnostics when enabled;
- and EAS Observe from Expo for app performance metrics and error reports with screen names, app version and device details.
For separately permitted Live talk, ElevenLabs transcribes recordings and creates reply audio; our backend sends conversation text, your learning status, exercise answers and a short summary of each pronunciation estimate to Anthropic (Claude Sonnet). When Ling checks how you say a phrase, Microsoft Azure Speech estimates your pronunciation from that recording. OpenAI also checks the text of vocabulary snapshots you explicitly confirm for sharing; Supabase handles publication, reports and creator blocks. For weekly league prizes, RevenueCat also receives your app user ID and the grant of Premium days.
In the iOS and Android apps, you can turn off usage analytics and error reports at any time in Data and privacy. In the web app, you can turn off usage analytics at any time in Settings. While it is on, your browser stores a random analytics ID. Turning it off deletes that ID and stops sending usage analytics from that browser. We share personal data with these services only for the purposes described in this policy. JulianDurian LLC is based in the United States.
Many of these providers also process data in the United States or in other countries outside the EU and UK. If you use ThaiHabit from the EU or UK, your data therefore leaves those regions. These countries may not offer the same level of data protection. When we share your data with these providers, we rely on the EU-US Data Privacy Framework and its UK Extension where a provider is certified. For other providers, we use the European Commission’s standard contractual clauses with the UK Addendum.
You can ask for a copy of these safeguards at support@thaihabit.com.
When usage analytics is enabled, ThaiHabit sends PostHog a signal when the mobile app comes to the foreground or the web app tab becomes visible, then about once a minute while the app stays in the foreground or the tab stays visible. The signal includes a random analytics ID and basic app information such as platform and version.
We use it to estimate how many app installations and browsers are open. Usage analytics starts off in the iOS and Android apps and starts on in the web app. You can change this in Settings.
Children’s privacy
ThaiHabit is for ages 13 and older. Where required for your Apple Account, we request Apple’s age range before access. A range below 13 blocks app use. You cannot override a required Apple age range with an in-app choice. ThaiHabit does not receive your birth date or identity document from Apple or store Apple’s age range in account records. ThaiHabit does not independently verify your age.
ThaiHabit has no parental consent process. If the law where you live requires a parent’s or guardian’s consent for a feature, including optional AI features, use it only with that consent. If you believe a child under 13 is using ThaiHabit, contact support@thaihabit.com so we can delete the data.
Data retention
Local app data remains on your device until you delete it, sign out of a linked account, delete that account, or uninstall the app. Account snapshots and other app-controlled account records are retained while the account is active and are removed through in-app account deletion, subject to bounded operational and legally required records. Deleting an account also ends any unused purchased Live minutes on it. They can’t be restored.
Generated audio for your own text is private to your account and is scheduled for deletion after 180 days without use, or when you delete your account. Cleanup removes the audio records and requests deletion of the stored files. A signed audio link already issued may remain usable until it expires, for at most one hour. Local audio downloads are cleared when you sign out or delete app data. Approved recordings from the general learning catalog are shared and can remain available independently of your account.
For the weekly league, daily XP totals are deleted after 90 days, sent league notifications after 30 days and weekly results after 12 months; your best league stays with your membership. Name reports and hidden learners are kept while both accounts exist. All league data is deleted when you delete your account.
AI response caches are deleted 30 days after last use, archived conversations after 180 days, network rate-limit records after 30 days, other rate-limit and import usage records after 90 days, AI quota records after 395 days and purchase webhook records after 730 days. Support requests, including your message and any email address you add, are deleted 12 months after we receive them. Store purchase history and provider security records follow Apple, Google, RevenueCat, hosting and AI/speech provider retention rules.
New league verification questions, associated XP awards and archived unverified open-week records are scheduled for deletion after 90 days, or earlier on account deletion. Regular cleanup processes a limited number per run, so deletion can occur later if a backlog remains.
Your rights
You can control local app data directly in the app, leave the weekly league in its settings, export linked-account data including league data and permanently delete a linked account from Profile. If you use a guest account, you can request a copy of your online data by email from Settings > Data and privacy.
The export lists the operational records it leaves out and why. You can ask us which reports concern your league name; we will not tell you who made them.
- View your learning progress within the app
- Reset all learning data at any time
- Delete all app data through the app settings
- Delete the app to remove local device data
- Leave the weekly league in its settings, delete a linked account and app-controlled cloud data in Profile, or contact support@thaihabit.com about applicable records outside those direct paths
App data deletion
You can delete all your app data at any time through the app’s menu. This will:
- Remove all learning progress
- Clear all app preferences and settings
- Clear the local Supabase session and detach the local RevenueCat identity where possible
- Reset the app to its initial state
This action requires confirmation and cannot be undone. It resets local app state. For a guest account, it also deletes the guest account and its server data. It does not remove App Store or Google Play purchase history or provider records. You can leave the weekly league in its settings without deleting private learning or your ThaiHabit account.
If you have a linked account, use Delete account in Profile to remove the Supabase Auth account, app-controlled cloud data and RevenueCat customer record. Apple-linked deletion reauthorizes with Apple and revokes the Apple authorization first. Records required for law, fraud prevention, security, or store purchase restoration may remain with the applicable provider; contact support for records outside the direct account path.
Subscription services
Premium subscriptions are purchased through Apple’s App Store or Google Play. RevenueCat helps verify purchase status and provide Premium access. When you make a purchase:
- Purchase information is processed by Apple’s App Store or Google Play
- RevenueCat verifies purchase status and provides Premium access
- We do not store your payment information
- Subscription status is stored locally on your device and mirrored to Supabase for AI entitlement limits
RevenueCat’s privacy policy can be found at: https://www.revenuecat.com/privacy
Speech processing
After separate Speech and pronunciation coaching permission, online pronunciation exercises send recorded audio and the practiced Thai reference text to our backend, which forwards them to Microsoft Azure Speech for transcription or pronunciation assessment and may forward audio to Hugging Face for ASR fallback or optional comparison. Our backend processes audio in memory for the request and does not store uploaded recordings. Separate Online reference audio permission allows the exact Thai text to be sent to ElevenLabs through our Supabase Edge Function.
Audio generated for your own text is stored privately for your account in Cloudflare R2 and delivered through short-lived signed URLs authorized by our Supabase backend. It follows the retention and account-deletion rules in section 6 and is not reused for other learners. Approved general learning recordings can be shared. Device speech is not used as a substitute for Thai recordings.
Contact us
ThaiHabit is provided by JulianDurian LLC, 1209 Mountain Road PL NE STE N, Albuquerque, NM 87110, USA. JulianDurian LLC is the controller responsible for the personal data described in this policy. Our representative in the European Union under Art. 27 GDPR is Julian Schubert, Aprikosenweg 42, 22175 Hamburg, Germany. If you have any questions about this Privacy Policy or our data practices, please contact us at:
Data protection rights
Depending on where you live, you can ask for access to your data, correction, deletion, restriction of processing or a portable copy. You can also object to processing based on legitimate interests. You have control over local learning data and direct export and deletion paths for linked-account data, including league data, through the app’s built-in features. You can leave the weekly league in its settings.
You can ask us which reports concern your league name; we will not tell you who made them. For questions about league reports, optional speech processing, subscription-related data, or provider-held records outside those paths, contact us at the address above. You also have the right to lodge a complaint with a data protection supervisory authority, in particular in the country where you live or work.
Legal bases for processing
If the GDPR or UK GDPR applies, we rely on these legal bases:
- Contract (Art. 6(1)(b)): your account, learning features, the weekly league, Premium access and support.
- Consent (Art. 6(1)(a)): optional AI, speech and audio features, usage analytics and error reports. You can withdraw consent at any time in Data and privacy. This does not affect earlier processing.
- Legitimate interests (Art. 6(1)(f)): security, abuse prevention and keeping the service reliable.
- Legal obligation (Art. 6(1)(c)): records we must keep by law.
Changes to this policy
We may update this Privacy Policy from time to time. Any changes will be reflected in the app with a new “Last updated” date. We encourage you to review this policy periodically.
Your consent
Using ThaiHabit is not blanket permission to share personal data with third-party AI. Before the first transfer for each optional tool, an in-app consent dialog identifies the exact data, purpose and possible providers: Anthropic, OpenAI, Hugging Face, Microsoft Azure Speech or ElevenLabs. Allow AI processing grants only the named purpose; permission for an image import does not unlock conversations, Live talk, speech, reference audio, answer comparison, live language analysis or another import type. Not now cancels that transfer and keeps the content on your device.
You can review or withdraw each permission at any time in Data and privacy; after withdrawal, ThaiHabit asks again before a new transfer for that purpose. Sharing a vocabulary set requires a separate confirmation for that exact snapshot. Its displayed set title, description, words and examples go through our backend to OpenAI for a safety check before an approved link is created; creating or editing a private set alone sends nothing for this check. Reports about shared vocabulary sets send the share link ID, reason and any details to ThaiHabit for review. They are scheduled for deletion after 365 days and removed during a regular cleanup run, or sooner when you delete your account.
Creator blocks hide that creator’s shared sets and stop Study Buddy connections; blocks remain until account deletion or authorized removal. Existing private sets, saved copies and audio are not deleted by reporting or blocking.
Website learning and download analytics
The website records visits to its landing pages and published learning guides, interactions with the sample lesson, the practice preview, guide exercises, print buttons and store buttons, and selected learning or plan options. In the practice preview, events record that you started, which step you finished and that you reached Daily Review; the keys you press are not recorded, and the preview does not use your microphone. Events retain only known page paths and their languages, including the first entry page, first landing page and most recently viewed guide.
A short browser-tab session connects these actions with a broad source category, such as Google, ChatGPT, an email campaign or an unknown referral. A visit from a social network also keeps that network’s name, such as TikTok or Instagram, and the short campaign and post codes from ThaiHabit’s own links. Store links carry short campaign codes from the link that brought you here, or otherwise the website placement, so Google Play and the App Store can report which source led to them.
The session uses sessionStorage and expires after 30 minutes without measured activity; if storage is unavailable, it lasts only within the current page. We do not create a persistent visitor identifier or include arbitrary page paths, raw referrer URLs, search terms, other campaign text, email addresses, newsletter tokens, click identifiers or free-text exercise answers in these event records.
Cloudflare processes these website events for aggregate reporting. The Analytics Engine dataset retains events for three months; operational logs may also record these events. Reports use separate landing-page and guide-visit counts to understand which visits lead to trying a Thai word, completing a guide exercise, requesting printing or opening a store. Each guide contributes at most one visit and one milestone of each kind per tab session; later store clicks are attributed to the most recently viewed guide. These counts do not establish a person’s identity, prove learning mastery, confirm a printed sheet or prove an app installation. This website measurement is separate from app analytics and newsletter subscription records.
For questions about website data, contact support@thaihabit.com. To estimate how many website tabs are currently visible, the website sends Cloudflare a random tab identifier when the tab is visible and approximately once a minute while it remains visible. The identifier is stored in the browser tab session and is not used to identify you or track you across devices. Cloudflare stores this identifier only temporarily for this count, stops counting it after five minutes without a signal and then deletes it automatically.
A tab that was closed or hidden may still be counted for up to five minutes. The signal does not include page paths or learning content.

