Privacy Policy
Last updated: November 12, 2026
Overview
Arabic Minute is an Arabic learning app. We respect your privacy and are committed to protecting it. We do not ask for your name, email address, or an account.
Data Stored on Your Device
The app stores the following data locally on your device. This data stays on your device except where noted under Online Features:
- Your selected difficulty level
- Words you have marked as favorites or learned, and your review schedule
- Your daily learning streak and XP progress
- Quiz and game high scores
- Notification time preferences
- Your conversation history with the app's tutor characters
- A random identifier generated by the app (see In-App Purchases below)
- In the free version, how many ads were shown today and when, and whether you have Premium, so that ads stay infrequent and Premium shows none
- Your ad privacy choices, which Google's consent tool stores on your device and Google's Mobile Ads SDK sends with each ad request (see Advertising), and a note of when the app first ran, so that ads wait until a later session
Online Features
Most of Arabic Minute works entirely offline. The following optional features require an internet connection and send data to our server, which forwards it to the processors listed under Third-Party Services.
- Text Conversation Practice: When you chat with a tutor character, the text you type is sent to our server to generate Arabic responses, hints, translations, and grammar corrections. We do not store conversation text on our server beyond what is needed to generate a response.
- Progress Summary: So that each tutor can pitch its Arabic to you, every chat message and every live call also carries a short summary of your progress: your level, your study streak, up to 40 of the words you have studied with their meanings, the words due for review and your favorites, how many messages you have exchanged with that tutor, and the last few topics you talked about. It is built on your device each time and is not stored on our server.
- Dictation (text chat): When you use the microphone in text chat, your speech is transcribed by your phone's speech recognition: Apple's on iPhone and iPad; on Android, your device's default speech recognition service, usually Google's. Depending on your device and language, that service may process the audio on its provider's servers under the provider's privacy policy; we never receive the audio. Only the transcribed text is sent to our server, as described above.
- Live Voice Conversations: While a live voice session is active, your microphone audio is streamed to ElevenLabs to run the spoken conversation. ElevenLabs turns your speech into text and passes that text, with the progress summary above, to an OpenAI language model that writes the tutor's replies. If ElevenLabs is unavailable, the call runs on OpenAI directly, which then receives your audio and the progress summary. During a call the app connects to ElevenLabs or OpenAI directly, so they also see your device's IP address. A transcript of the session is kept on your device. What ElevenLabs and our server keep is described under Data Retention and Deletion.
- Grammar Notes After a Call: When a live call ends, the lines you spoke, as transcribed during the call, are sent to our server and on to Anthropic to write the grammar notes in the call recap.
- Reporting a Reply: When you report something a tutor said, in a chat, during a live call or from the call recap, the app sends the reported line (up to 2,000 characters), the reason you chose, your note if you write one, the app version and whether your device runs iOS or Android to our server. It is kept for the developer to review (see Data Retention and Deletion) and is not sent to any AI service. Nothing else from the conversation is sent, and no device identifier is stored with the report.
- Ask AI (Dictionary): When you tap Ask AI, the word or phrase you typed is sent to our server and on to Anthropic for a translation. If you play the word, its audio is made as described under Text-to-Speech.
- Text-to-Speech: For words and phrases without bundled audio, including words you play from Ask AI, for the tutors' replies, and for corrected versions of your sentences, the app sends the text to our server, which generates the audio with ElevenLabs or Microsoft. Our server keeps a copy of the audio for a limited time so that repeat plays are quick (see Data Retention and Deletion). If the server is unavailable, the app falls back to your device's built-in speech synthesis.
Your Choice
From version 1.2.1, Talk (the tutor chats and live calls) and Ask AI in the Dictionary ask before they send anything. The notice names what is sent and to whom, and nothing is sent until you agree. To withdraw, turn off Settings > Talk and AI, which is also in the app from version 1.2.1; nothing more is sent until you agree again. Earlier versions do not ask first: they send data as described above when you use these features. The rest of the app, including word audio, works without agreeing.
In-App Purchases
Arabic Minute offers an optional Premium subscription, billed by Apple through your Apple ID in the iOS app, or by Google Play through your Google account in the Android app. We never receive your name, email address, or payment details. To enforce the daily live-voice and chat allowances, our server stores a random app-generated device identifier together with the live-voice seconds and the number of AI requests (chat messages, and translations, corrections and hints) used each day and, for subscribers, the Apple transaction identifier of the subscription or, for a Google Play subscription, a one-way hash of its purchase token. This identifier is not linked to your identity and is not used for tracking. From version 1.2.1 it stays between the app and our server: it is not passed to ElevenLabs, OpenAI, Anthropic, Microsoft, or Google, and each live call carries a new random label instead. Version 1.2.0 also attached it to live calls on ElevenLabs as the session's user ID; update the app to stop this.
In the Android app, a purchase also gives Google Play a one-way hash of the identifier, never the identifier itself, as the purchase's account reference, and our server sends Google Play the purchase token to confirm the purchase before it unlocks Premium.
Advertising
From version 1.3.0, the free version of the app shows a short full-screen ad from Google AdMob between some activities, for example after you finish a lesson or a quiz, or close the recap of a call. Ads never appear during a lesson, a chat or a live call, never in the first session after you install the app, and never when you are offline. Premium removes ads.
- Non-personalised ads only: the app asks Google only for non-personalised ads, which are not based on your past activity. Non-personalised ads may still be chosen using the app's content and your general location. We do not track you across other companies' apps or websites. On iPhone and iPad the app never asks for permission to track (Apple's App Tracking Transparency), so it does not give Google your device's advertising identifier (IDFA). On Android, Google's Mobile Ads SDK may read your device's advertising ID, which you can reset or delete in your device's settings.
- What Google may collect: to show ads, limit how often they appear, measure them and prevent fraud, Google's Mobile Ads SDK may collect device identifiers such as the identifier for vendors (IDFV) on iPhone and iPad, or the advertising ID and app set ID on Android, your IP address (which may be used to estimate your general location), information about the ads you see and how you use the app (for example app launches and taps on ads), and crash and performance data such as launch time. Google processes this data as an independent controller under the Google Ads Controller-Controller Data Protection Terms and its privacy policy, which protect it at least as well as this policy describes; see How Google uses information from sites or apps that use its services.
- Other ad companies: Google chooses each ad in an auction. Through it, other advertising companies (the ad partners listed in Google's consent form) may receive information such as your device type, your approximate location and information about the app, to bid on and show an ad, under their own privacy policies.
- Consent: before showing ads, the app asks Google's consent tool whether a consent message applies to you, which sends Google your IP address and basic information about your device and the app. In the European Economic Area, the United Kingdom and Switzerland, Google's consent form asks for your choices before any ad is requested, and the app requests ads only when Google's consent rules allow it. If you decline, Google may still show limited ads, which do not use identifiers stored on your device for advertising. Where privacy laws in some US states call for it, Google's message also lets you opt out.
- Changing your choices: where Google's consent rules require it, Settings > Ads > Privacy choices opens Google's form so you can review or change your choices at any time. You can also manage how Google uses data for ads in My Ad Center, and see or delete data Google holds with the tools described in Google's privacy policy. Premium removes ads and stops the app's ad requests altogether.
- Reporting an ad: if an ad seems inappropriate, Settings > Ads > Report an ad writes us an email with the app version and when the last ad appeared, so that we can block it.
- What we receive: Google does not give us data about you. We see only aggregated reports of how ads perform and what they earn.
Ad Measurement (Meta)
From version 1.2.5, the iOS app (iPhone and iPad) includes Meta's Facebook SDK so that we can count how many people install the app after seeing our ads on Instagram and Facebook. Each time you open the app, the SDK sends Meta an "app opened" event (the first one marks the install) with technical details about your device, such as its model, operating system version, language, time zone and the app's version, and a random identifier the SDK creates on your device. Meta also sees your device's IP address. We have switched off the SDK's automatic event logging, its collection of Apple's advertising identifier (IDFA) and its optional extra event details, so Meta does not receive what you study, type or say, and the app does not ask to track you. Meta uses this data to measure our ads' results and to run and improve its ads services, under the Meta Privacy Policy, and keeps it under its own retention rules; we cannot delete it from Meta's servers. The app also registers its install with Apple's SKAdNetwork, which tells the ad network an ad led to an install without identifying you or your device. The Android app does not include the Facebook SDK and does not send these events.
Data Retention and Deletion
Our server keeps only the following:
- Live-voice minutes: the device identifier with the seconds of live voice and the number of live calls used that day, kept for about two days (50 hours) and then deleted automatically. A marker that stops the same call from being counted twice is kept for the same time.
- AI request counts: the device identifier with the number of chat messages, of translations, corrections and hints, and of characters spoken in a tutor's own voice used that day. To limit abuse, the server also counts these requests and live calls per network under a keyed hash of your IP address that changes every day and cannot be reversed without our server's secret key; the address itself is not stored. Both are kept for about two days (50 hours) and then deleted automatically.
- Premium: for subscribers, a record that the device identifier has Premium, kept until the current subscription period ends (for a Google Play subscription that renews automatically, up to two days longer, so that a renewal confirmed late does not interrupt Premium), and a record linking the subscription's Apple transaction identifier (for a Google Play subscription, a one-way hash of its purchase token) to the identifiers of up to three devices that use it, kept until the period ends or for 30 days, whichever is longer. A renewal or a restore writes them again.
- Generated speech: audio our server has generated, stored under a one-way hash of the text and the voice, for 7 days (Microsoft Edge voices) or 30 days (ElevenLabs and Azure voices), then deleted automatically. It holds no device identifier or anything else that identifies you.
- Reports: a reply you report, with your reason, your note, the app version and the platform, kept for 90 days so the developer can review it, then deleted automatically. It holds no device identifier or IP address, so please leave personal details out of the note. To stop one device or network from sending many reports, the number sent each day is counted per device identifier and per network, like the AI request counts above, and those counts are kept for about two days (50 hours).
Conversation text and the progress summary pass through our server only to reach the services that process them, and are not stored on it, except a reply you choose to report (see Reports above). Call audio goes straight from your device to ElevenLabs or OpenAI and never reaches our server. Since October 2026, our ElevenLabs settings, for calls from any version of the app, tell ElevenLabs not to record call audio and to delete call transcripts shortly after each call. Calls made before that change remain under ElevenLabs' standard retention settings. Your conversation history and progress stay on your device, and deleting the app deletes them.
Live-voice and AI request records delete themselves within about two days, reports after 90 days, and Premium records end with your subscription. To ask us to delete data sooner, or with any question about what we keep, email us at the address under Contact.
Third-Party Services
- Cloudflare: hosts our server, which relays the online features above, and the storage described under Data Retention and Deletion.
- Anthropic: its Claude models write chat replies, hints, translations, Ask AI lookups, and grammar notes, including notes on a live call after it ends, from the text you send and, in Talk, your progress summary. Under Anthropic's commercial terms, data sent through its API is not used to train its models by default, and Anthropic deletes it within 30 days unless it must keep it longer to meet a legal requirement or to enforce its usage policy.
- ElevenLabs: runs live voice conversations and generates spoken audio for the tutors and for words. For spoken audio it receives only the text to be spoken, with no identifier. We have opted out of ElevenLabs using this content to train or improve its models. ElevenLabs states that its agreements with the language-model providers it uses prohibit them from training their models on this content.
- OpenAI: provides the language model that writes the tutor's replies during live calls on ElevenLabs, and runs live calls directly when ElevenLabs is unavailable. On calls through ElevenLabs, ElevenLabs' agreement with the model provider applies (see ElevenLabs above). When a call runs on OpenAI directly, data sent to OpenAI's API is not used to train its models by default, and OpenAI keeps it in abuse-monitoring logs for up to 30 days, longer only where the law requires it or to protect its services or others from harm.
- Microsoft: neural voices (Azure Speech and Microsoft Edge speech) for the tutors' speech, corrected sentences, and words without bundled audio. Only the text to be spoken is sent, with no identifier.
- Apple: handles in-app purchases in the iOS app, and the speech recognition for dictation on iPhone and iPad, which runs on your device or on Apple's servers depending on your device and language.
- Google: Google Play handles in-app purchases in the Android app; our server sends it the purchase token to confirm and acknowledge each purchase. In the Android app, Google's speech recognition usually also handles dictation (when it is your device's default speech recognition service), on your device or on Google's servers depending on your device and language.
- Meta: receives the app-opened and install events from the Facebook SDK in the iOS app described under Ad Measurement, to measure installs from our ads.
- Google Fonts: the app loads fonts from Google Fonts for display purposes.
- Google AdMob: shows the ads in the free version and runs the consent form described under Advertising, under Google's privacy policy.
Anthropic, ElevenLabs and OpenAI process this data only to provide their services to us, under terms that protect it at least as well as this policy describes. Microsoft receives only the text to be spoken, with no identifier. Google processes the data its Mobile Ads SDK collects as an independent controller, under terms that protect it at least as well as this policy describes (see Advertising). Meta processes the install and app-open events its SDK reports only to measure our own ads (see Ad Measurement).
Analytics
Arabic Minute has no analytics of its own: the app does not send how you use it to an analytics service, and we do not track you across other companies' apps or websites. In the free version, Google may use the ad data described under Advertising for analytics about ad delivery and performance; we see only aggregated reports of it. In the iOS app, Meta's SDK reports app installs and app opens so we can measure our own ads (see Ad Measurement); it does not collect the advertising identifier.
Children's Privacy
The app does not knowingly collect personal information from children under the age of 13. Conversation text and voice audio are processed in real time to generate a response and are not stored on our server, except a reply someone chooses to report. The install and app-opened events described under Ad Measurement are sent for every user of the iOS app; they contain no name, contact details, or learning content.
Changes
If we update this policy, we will post the revised version here with an updated date.
Contact
If you have questions about this privacy policy, please contact us at abdulla.saeed.alhemeiri@gmail.com.