Privacy Policy — Senthora
Last Updated: 29 October 2026
Data Controller: Intersolution Trade OOD (official Latin form: Intersolution Trade LTD)
Registration Number (UIC/EIK): 208418306
Address: 2B Topli Dol Str., fl. 4, ap. 17, Krasno Selo district, 1618 Sofia, Bulgaria
Email: support@senthora.ai
Web: https://senthora.ai
1. OVERVIEW
Intersolution Trade OOD ("we", "us" or "our") is the operator of Senthora, an AI-native language tutor. This policy explains what personal data we process, why, who receives it, and your rights.
Questions about your data? Email support@senthora.ai.
2. WHO PROCESSES YOUR DATA
To provide real-time tutoring, Senthora sends data to the third-party providers named below.
Providers that receive your voice or conversation content
- LiveKit, Inc. (USA): Carries the live audio between the app and the tutor, including noise cancellation and detecting when you have finished speaking.
- AssemblyAI, Inc. (USA, processed on its EU servers): Turns your speech into text during voice sessions.
- OpenAI, Inc. (USA): Writes the tutor's replies from the conversation text, and produces your session summary and your grammar and vocabulary notes. It receives text only, never your voice. OpenAI also receives a pseudonymous identifier for its own abuse-prevention purposes, for which OpenAI acts as an independent controller. It may keep abuse-monitoring logs for up to 30 days.
- ElevenLabs, Inc. (USA): Turns the tutor's written reply into the voice you hear. It receives the text of the tutor's reply, never your voice, and no user or session identifier is sent with it. Under its terms, ElevenLabs may keep request data for a limited time for security and troubleshooting.
- Microsoft Azure Speech (EU region): Scores your pronunciation.
Your voice goes to LiveKit, AssemblyAI and Microsoft. OpenAI and ElevenLabs receive text only.
Infrastructure and operational providers
- Amazon Web Services (EU — Frankfurt): Application hosting and file storage, including profile pictures.
- MongoDB Atlas (EU — Frankfurt): Database hosting for your account, voice session records and learning data.
- Google Firebase: Sign-in, push notifications, crash reporting (Crashlytics) and, only if you allow analytics, product analytics.
- Sentry: Error monitoring, configured not to collect personal data.
- RevenueCat (USA): Subscription and purchase verification. If you allow analytics (and, on iPhone/iPad, tracking), it also passes purchase events to Meta and AppsFlyer (see below).
- Apple App Store / Google Play: Process payment for subscriptions. We receive your purchase status, through RevenueCat. We never receive your card details. Apple and Google handle payment as independent controllers, under their own privacy policies.
Advertising measurement (only if you allow analytics)
- AppsFlyer: Marketing attribution, which tells us which ads lead to installs and trials. It runs only if you have allowed analytics. On iPhone/iPad, if you do not allow tracking, it runs without your device's advertising identifier (see 3.D). If you have allowed analytics and, on iPhone/iPad, tracking, we also send purchase events (such as starting a trial) to AppsFlyer through RevenueCat.
- Meta Platforms (Facebook SDK and Conversions API): If you allow analytics, we send limited app events to Meta through its SDK, and purchase events (such as starting a trial or subscribing) to Meta's Conversions API through RevenueCat. On iPhone/iPad, if you do not allow tracking, Meta's SDK runs in its limited mode (see 3.D) and we send Meta no purchase events. If you do not allow analytics, we send Meta no app events and no purchase events. Meta and we are joint controllers for this advertising measurement (GDPR Art. 26). In short: we are responsible for telling you about it and for asking your permission; Meta is responsible for what it does with the data once it receives it. You can contact either of us about it. See the Meta Privacy Policy.
International transfers: Where data goes outside the EEA, we rely on Standard Contractual Clauses and, for providers certified under it, the EU–US Data Privacy Framework. Contact us for a copy of the safeguards.
Training of AI models
We do not use your conversations, text prompts or voice recordings to train AI models. The providers that process them either state the same for themselves, or we have opted out of model training in our account with them:
- OpenAI states that data sent to its API is not used to train or improve its models. We have not opted in.
- Microsoft states that it does not use customer data to improve its Speech models, and that it does not retain or store the data sent for pronunciation assessment.
- LiveKit states that it does not use customer content, prompts, transcripts or audio to train or fine-tune its models, and that it does not retain media.
- AssemblyAI: we use its European servers and have opted out of model training.
- ElevenLabs: we have opted out of model training in our account settings. It receives only the text of the tutor's replies.
3. TYPES OF DATA WE COLLECT
A. Account and identification data
Your name, email address, sign-in identifiers from Apple or Google, your profile picture if you add one, and your account preferences. We use them to run your account and keep your learning progress across devices.
B. Voice and speech data
When you use voice features (voice sessions with the tutor, or pronunciation practice), your speech is processed to turn it into text, score your pronunciation and produce the tutor's replies.
- How we use it: Senthora does not create or store a voiceprint and does not identify you by your voice. Your voice is shared only with the processors that handle it for the session (section 2), and never for advertising.
- Who handles it: LiveKit carries the live audio, AssemblyAI turns it into text, and Microsoft Azure Speech scores your pronunciation. OpenAI and ElevenLabs receive text only.
- Retention: Senthora does not store your audio. We keep only a record that a session took place and how long it lasted (see section 6).
C. Learning history and transcripts
We store written transcripts of your conversations with the tutor and the learning data that comes from them (vocabulary, grammar notes, pronunciation scores, session summaries) to give you feedback and to track your daily practice.
Please don't share health details, political or religious views, or anything else you'd consider sensitive with your tutor. You don't need to in order to practise. We don't detect, categorise or target anything about you based on what you say.
D. Technical, analytics and advertising measurement data
Your analytics choice: When you first open the app, before you sign up, we ask whether you want to help improve Senthora. It stays off unless you choose Allow. If you allow it, we use Firebase Analytics for product usage events, AppsFlyer for attribution, and the Meta SDK and Meta Conversions API for advertising measurement (section 2). If you don't, none of these run. You can change your choice at any time in Settings → Help improve Senthora. If you switch it off, we stop sending this data from then on.
iOS App Tracking Transparency (ATT): On iPhone/iPad, if you allow analytics, Apple's tracking prompt follows. If you choose Allow, we may use the device advertising identifier (IDFA) for attribution with AppsFlyer and Meta, and purchase events are sent to Meta and AppsFlyer as described in section 2. If you choose Ask App Not to Track, we do not use your device's advertising identifier (IDFA), and we do not send your email, phone number or name to Meta or AppsFlyer, from your device or through RevenueCat. Meta's SDK still runs in its limited mode: it reports basic app events, such as installing and opening the app, together with basic device information, a random ID created for this installation and, as with any internet connection, your IP address. Meta says it removes identifiers and combines this data across users before using it to measure ads (Aggregated Event Measurement). Apple's SKAdNetwork also tells Meta how many installs an ad produced, without saying who they are. AppsFlyer still runs, without your device's advertising identifier, and receives no purchase events. Product analytics (Firebase) can still run if you allowed analytics. You can change this at any time in iOS Settings → Privacy & Security → Tracking.
Android advertising ID: On Android, if you allow analytics, the Google advertising ID may be used for attribution with AppsFlyer and Meta.
Not affected by your analytics choice: Crash reports (Firebase Crashlytics), error reports (Sentry) and the device token used to deliver push notifications (Firebase Cloud Messaging) do not depend on your analytics choice, because they keep the app working and deliver notifications. Subscription processing through RevenueCat and the app stores is needed to deliver purchases you make.
Notifications: Service notifications are part of the app. Practice reminders and news are optional and are sent only if you switch them on; you can switch them off at any time in Settings.
IP addresses may be processed by infrastructure providers as part of normal network requests.
E. Why we use your data (legal bases)
- Performance of our contract with you (Art. 6(1)(b)): your account and sign-in, voice and speech processing for tutoring and pronunciation scoring, conversation transcripts, your learning record and practice history, service notifications, and your subscription and purchases.
- Your consent (Art. 6(1)(a)): product analytics, advertising measurement and attribution (on iPhone also your tracking choice), and reminder or news notifications. You can withdraw consent at any time (section 6).
- Legal obligation (Art. 6(1)(c)): keeping purchase and accounting records.
- Our legitimate interests (Art. 6(1)(f)): crash and error reports, to keep the app working, and abuse prevention with OpenAI (section 2).
What we need from you: sign-in details are needed to create an account, and your date of birth is needed once, to check your age before voice features. Without them we cannot provide those features. Analytics, tracking and reminder or news notifications are optional, and the app works without them.
4. AI TRANSPARENCY
- AI identification: Your tutor is an AI, not a human. The app tells you so.
- Accuracy: AI systems can make mistakes and may give incorrect language feedback. Senthora is a learning aid, not a certified translation, legal or medical service.
- Right to Flag: You can report any AI response from the app (on each AI message and on the post-call summary). Reports are sent to us, and you receive an acknowledgement by email.
- Synthetic voice: The tutor's voice is AI-generated. We do not currently mark synthetic audio with a machine-readable label, because we do not have a way to apply one.
5. PROTECTION OF MINORS (16+)
Senthora is for users aged 16 or older. Before your first voice session we ask for your date of birth. We use it only to check your age at that moment: we do not store it, and keep only the fact that the check was passed and when.
If the check shows you are under 16, you cannot use Senthora. The app stays blocked, and we keep a note on that device that the check was not passed, so it cannot be retried there. This applies to the installation of the app you used. We do not knowingly process voice recordings from anyone under 16. If you believe someone under 16 has created an account, email support@senthora.ai and we will delete it. The same applies if you were blocked by the age check and want your account deleted.
6. DATA RETENTION AND YOUR RIGHTS
How long we keep your data
- Account details (name, email address, sign-in identifiers) — while you have an account; deleted when you delete it.
- Written transcripts of your conversations — 12 months.
- Your learning record (vocabulary, grammar notes, pronunciation scores, session summaries, practice history) — 24 months after you last use Senthora. "Last use" means the last day you opened the app or had a session.
- Voice session records (when a session happened and how long it lasted, not the audio) — kept for 24 months after the session, and deleted earlier if you delete your account.
- Operational logs, including notification delivery logs — 90 days.
- Crash reports (Firebase Crashlytics) — 90 days.
- Product analytics (Firebase) and error reports (Sentry) — kept only for a limited period set in the provider's account settings.
- Attribution data at AppsFlyer — up to 24 months, under AppsFlyer's terms. Advertising data at Meta — up to 2 years, under Meta's terms.
- Abuse-monitoring logs at OpenAI — up to 30 days, under OpenAI's terms (section 2).
- Purchase records — while you have an account. After you delete it, we keep them for 5 years in pseudonymised form (transaction facts only; identifiers that would re-attach them to you are removed), because tax law requires it.
Your choices and rights
- Right to Delete: Use the "Delete Account" button in App Settings. This deletes your account and learning data from our systems. Purchase records are kept for 5 years in pseudonymised form (identifiers that would re-attach them to you are removed), because tax law requires it. Deletion requests are also sent to RevenueCat and AppsFlyer. Meta, as joint controller, handles deletion of its own copy; you can also ask Meta directly. We do not currently keep database backups, so deleted data is not kept in backup copies. Deleting your account does not cancel a subscription bought through Apple or Google — you must cancel it yourself in your App Store or Google Play account settings.
- Right to Export: You can download a machine-readable copy of your data from Settings.
- Voice permission: Before your first voice session we ask for two separate permissions: to record, and to share your voice with the voice processors in section 2. Both start unticked. If you decline, your account stays and voice features stay locked. You can withdraw these permissions at any time in Settings → Privacy → Voice recording permission. Withdrawal stops future voice sessions; it does not by itself delete past data (use Delete Account for that).
- Withdrawing consent: You can switch analytics off in Settings → Help improve Senthora, change tracking in iOS Settings → Privacy & Security → Tracking, and switch reminders off in Settings. Withdrawing does not affect what was done before.
- Your rights: You have the right to access, rectify, erase, restrict or port your data, and to object to processing based on our legitimate interests. Email support@senthora.ai; we reply within one month. You may also lodge a complaint with the Bulgarian Commission for Personal Data Protection (cpdp.bg) or with the supervisory authority where you live.
7. DATA SECURITY
- In Transit: Data is encrypted in transit (TLS 1.3).
- At Rest: Data is encrypted at rest (AES-256). Application hosting and file storage (including profile pictures) run on Amazon Web Services in the EU (Frankfurt), and our database runs on MongoDB Atlas in the EU (Frankfurt). For transfers outside the EEA, see section 2.
8. CHANGES TO THIS POLICY
When we change this policy, we update this page and the "Last Updated" date at the top.
9. CONTACT INFORMATION
Intersolution Trade OOD (official Latin form: Intersolution Trade LTD)
UIC/EIK: 208418306
2B Topli Dol Str., fl. 4, ap. 17, Krasno Selo district, 1618 Sofia, Bulgaria
Email: support@senthora.ai