Parlette Privacy Policy

Version 1.2 of 20 October 2026

We send you previous versions on request: write to support@parlette.app.

This policy applies to the Parlette app (iPhone and Android) and to the website parlette.app. It explains what personal data we process and why, who we entrust it to, how long we keep it and what rights you have.

In short

• Parlette belongs to Univapp Srl, based in Bari. For any question, write to support@parlette.app.
• We process only the data the app needs: your account, what you study, the words you save, your conversations with the AI and the status of your subscription.
• The database with your data is in Frankfurt, in the European Union. AI, voice and subscriptions go through providers in the United States, with the safeguards required by the GDPR (point 7).
• The text of conversations is processed by Claude, Anthropic's AI. Voice is processed by ElevenLabs. Neither of them can use your data to train their own models.
• We don't save recordings of your voice and we don't keep the text of conversations on our servers. The AI providers also keep data for a limited time, set out in point 8.
• The AI can make mistakes. You can turn off the AI whenever you like, in Profile (point 5).
• Payments are handled by Apple and Google. We never see your card details.
• We don't show ads, we don't track you and we don't sell your data. The website doesn't use cookies.
• If you want, we collect usage stats without your name with PostHog, in the EU. They start only if you say yes, and you can turn them off in Profile.
• Parlette is for people aged 18 and over.
• You can see your data, correct it, download it, delete it and object to some uses. You delete your account from the app, in Profile > Delete account. You can also lodge a complaint with the Italian data protection authority (Garante privacy).

1. Who we are and how to contact us

The data controller is whoever decides how and why your data is used. For Parlette it is:

• Univapp Srl
• Via Orfeo Mazzitelli 140, 70124 Bari (Italy)
• VAT no. 09131160724
• REA BA-673911
• PEC (certified email) univappsrl@pec.it

For privacy questions and to exercise your rights, write to support@parlette.app.

We don't have a data protection officer (DPO), because for our activity the law doesn't require one (art. 37 GDPR).

2. What data we process and why

Below you'll find the data divided by feature. The legal bases are in point 3, the providers in point 6, the retention periods in point 8.

Where the data comes from

You give us almost all the data while you use the app. Some comes from others:

• if you sign in with Apple or Google, they send us your email, name and an identifier;
• the status of your subscription comes from the App Store and Google Play, through RevenueCat.

We take your native language from the language set on your phone. You can change it in your Profile.

2.1 Account and sign-in

To create your account and let you sign in we process:

• your email. If you use Apple's "Hide My Email" we receive an address created by Apple, which then forwards our messages to you;
• your first name, if Apple or Google sends it to us or if you type it yourself. In the Challenge with friends it's the name other players see (point 2.5). You can change it in your Profile;
• an internal identifier for your account, the identifier Apple or Google gives you and the method you use to sign in;
• the 6-digit code we send you by email when you sign in with your email. It expires after 10 minutes (point 8);
• session data: the temporary codes that keep you signed in, your IP address and the type of device and app. These also go into the access log;
• your native language;
• the version of the Terms you accepted, with date and time. You accept them by tapping "Continue" when the app asks for your year of birth, and in doing so you also declare that you are at least 18 (point 11);
• the version of the AI line, with the date and time you tapped "I agree" (point 5).

Apple sends us your name only the first time you sign in. Google sends us your full name and the link to your profile photo: we keep only your first name, and we remove your surname and photo from the sign-in data straight away. The technical logs of our provider may contain your full name for 1 day only (1 hour for sign-in logs), then they delete themselves.

If you sign in with Apple, we keep an encrypted code from Apple that is used only to revoke your access at Apple when you delete the account, as Apple requires. Only our server reads it, never the app, and it is deleted together with the account.

If someone requests a code with your email and doesn't use it, we delete the address within 24 hours.

Parlette doesn't use passwords.

We also use your email for service messages: the sign-in code, the contract confirmation after the purchase, the notice at least 30 days before your annual subscription renews, the withdrawal receipt, support replies and the notice before we delete an inactive account (point 8). We don't send you marketing emails.

2.2 Study, preferences and saved words

For your study plan we process the languages you study, the level of each one (from A1 to B2), the active language, the "I know this" ticks on grammar rules, the voice you chose for each language and your plan.

At first launch we ask you about your level, goal, why you're learning, interests and AI style. The answers stay on your phone. Only one line with your choices goes to our server, always picked from a list: level per language, reason, goal, minutes per day, interests, AI style, how much to correct you and weak points. We send it to the AI together with the conversation, to adapt the replies.

For each word you save we keep:

• the word and its language;
• the base form and two translations;
• the "to review" mark;
• the date;
• how you saved it (by voice, by typing it, from the pop-up, from a conversation or from the Challenge).

Words stay until you delete them (swipe on the word) or delete your account. How many words you can save on each plan is in the Terms. If you go back to Free, the words you have already saved stay.

Pop-up translations, flashcard example sentences, Challenge words and pronunciation audio for common words are the same for everyone. They contain no data about you. When you look up a word the dictionary doesn't have yet, we send it to Claude together with the code derived from your account (point 5); only the word stays in the dictionary, without who looked it up. Words you find in the downloaded dictionary are looked up by your phone, without asking the server anything.

You choose your level, and "I know this" is a tick you add yourself. No system assesses you, except in the optional level test (point 5).

If you take the level test, which is optional, for each test we keep the language, the level, the score of each part and the date. Your written answers and the transcript of what you say go to the AI to be scored, as in the other AI features (point 5), and we don't save them.

2.3 Conversations with the AI

When you speak or write with the AI (chat, call, vocabulary building, Practice and Review with "My words") we process:

• your messages (at most 30 words each), the AI's replies and corrections;
• the written transcript of voice notes and calls, both yours and the AI's;
• the topic or theme you chose and the new words the AI suggests at the end of the conversation.

We need them to let you have conversations, to correct you and to offer you "Save all". Written messages and voice notes pass through our server in Frankfurt only to reach the AI, and we don't save them; for calls see point 2.4. To show you the last 3 conversations for each language in Speak, your phone remembers only the topic, the mode and the AI's last sentence. What we send to the AI is explained in point 5.

2.4 Voice and microphone

The app uses the microphone in these features:

• Save word (from the widget, the fixed microphone or the app icon);
• the first-launch Try it;
• voice notes in chat;
• the call and spoken vocabulary building;
• Practice and Review by voice;
• the speaking part of the level test;
• the Challenge in Voice mode.

Before your phone asks for permission, the app explains in one line what the microphone is for, and it always shows you when it is listening. You can remove the permission whenever you like from your phone's settings. Without the microphone you can use Parlette in writing: chat, words typed or saved from the pop-up, the Challenge in Text mode.

The microphone hears your voice, background noise and sometimes the voices of people near you. Keep this in mind when you're not alone.

The path your voice takes depends on the feature.

• Save word, voice notes and Try it. Your phone sends the audio to our server in Frankfurt, which passes it to ElevenLabs to transcribe it: ElevenLabs doesn't see your IP address. We don't save the audio anywhere, and as soon as the transcript arrives we ask ElevenLabs to delete it. Outside Try it, the transcribed text then goes to Claude, which finds the base form and translations of the word, or replies to you in chat.
• Real-time voice conversations (call, spoken vocabulary building, Practice and Review by voice). Your phone connects directly to ElevenLabs with a temporary access code that our server gives it, so ElevenLabs also sees your IP address. In a call the replies are written by Claude, which ElevenLabs uses through its providers, Google Cloud (Vertex AI) or Anthropic. ElevenLabs data may be located in the United States, the EU or Singapore. We keep neither the audio nor the transcript: after each call we ask ElevenLabs to delete it. After the call, ElevenLabs creates a short automatic summary with one of its own models; we delete it together with the call.
• Turn-based voice conversations. They follow the voice note path: your phone sends the audio to our server, ElevenLabs transcribes it and creates the voice for the replies, and the replies are written by Claude through our server. ElevenLabs doesn't see your IP address, and we delete the transcripts as soon as they arrive, as for voice notes.
• Replies read aloud. ElevenLabs receives the text of the reply and creates the synthetic voice; straight afterwards we ask for it to be deleted, and we don't keep the audio of the replies. The pronunciation of the words you save or tap is also created by ElevenLabs. The file stays only on your phone, until you sign out or delete it, and we don't keep it on our servers.
• Challenge by voice: see point 2.5.

In Try it, your account doesn't exist yet. The audio goes from our server to ElevenLabs without any identifier, and nothing goes to Claude. To prevent abuse we count trials with an encrypted code derived from your network: we don't save the IP address, and we delete the code within 2 days. The languages you chose and your first word stay on your phone and go into your account when you sign in.

Univapp never saves recordings of your voice. How long ElevenLabs keeps them is in point 8.

We don't create a voiceprint, we don't recognise who is speaking, we don't clone your voice and we don't analyse emotions or health. We need your voice only to understand what you say, so we don't treat it as biometric data (art. 9 GDPR). ElevenLabs processes the audio on our behalf, and in our account the use of content to train its models is turned off.

2.5 Challenge and camera

The Challenge has three modes:

• Solo: you practise alone against the clock. The result and your best score stay on your phone;
• Friend: a live game for 2 to 8 people, each on their own phone. The person who sets it up shows a QR code and the others scan it;
• Random: the server matches you with a player who is studying the same language at the same level. If no one is found within 25 seconds, you can practise alone, without points. Opponents are always people: in the Challenge you don't play against the AI.

To make the game work we process:

• the session code. The QR contains only a link with a random code, which expires after 10 minutes;
• the game settings: language, level, Voice or Text, length;
• the name other players see and your points. In Friend mode it's your Profile name, with the initial in the avatar. In Random and in the leaderboard it's your player name, which you make up or the app suggests. Others don't see your email, your account, your photo, your plan or your native language;
• what happens during the game: the word in play, the time, who gets the point, the points, who leaves ("Marco left") and the rematch.

Every name goes through a list of banned words, in the app's 5 languages, so that nobody uses offensive names.

Your phone checks every answer: wrong ones stay there. The right one goes to the game server (Supabase, in Frankfurt), which checks it again with fixed rules, gives the point to whoever is first and doesn't save it: it keeps only who got the point. In Voice mode your phone sends an answer only if it heard it loudly, meaning said by you, and it doesn't send the volume to anyone. In the Challenge no answer goes to the AI. During a game only words and answers pass between players: there is no chat.

Your native language is used by your phone to show you the word in your language. The game server deletes it with the game, and it doesn't reach other players. You save the words you didn't guess with "Save all", and they go into your words (point 2.2).

In Random mode and in the leagues you play with people you don't know, chosen by the server among those studying the same language at the same level. To do this the server keeps the queue of people looking for a game, game results and the week's points. Each week the points from your Random games go into your league's leaderboard, with the Bronze, Silver and Gold tiers. In your group's leaderboard, up to 30 people see your player name and your points. If you reach the top 100 of the league, all players in the league see them. They never see your email, your account or your Profile name. The "Parlette team" players you see in the leaderboard are simulated players: they are not people and they are nobody's data.

You can report an offensive name by pressing and holding it in the leaderboard. If 3 people report it, the name is hidden: others see "Player" with a number, until the person who chose it picks a new one.

To play with other people you need a Parlette account. The data of a game with friends is deleted within one day of its creation. League results and points are kept for at most 3 months (point 8).

While it's valid, the QR code lets in anyone who scans it, even from a photo. Show it only to the people you want to play with.

The camera is used only to read the QR code. Your phone reads the image, doesn't save it and doesn't send it to anyone. On Android the QR code is read by ML Kit, a Google library that works on the phone, and the image doesn't leave it. ML Kit does, however, send Google some technical data: phone model and operating system, performance, error codes and an installation identifier. Google uses them to measure and improve the library and to detect abuse. If you don't give camera permission, you can scan the code with your phone's Camera app: the link opens the Challenge in Parlette.

In Voice mode the microphone listens for the whole game and can also hear the other players. Speech recognition is done by the phone's speech service and Parlette doesn't send the audio to anyone: how the system handles it depends on your phone's settings and on Apple's or Google's privacy policies. If your phone can't recognise words in the language of the game, you type your answers and the others carry on by voice.

If someone scans the QR without having the app, the page parlette.app/sfida/… opens and then the store. In that case the website rules apply (point 2.11).

2.6 Reports

In every conversation with the AI there is "Report": in the "…" menu at the top, or by pressing and holding an AI reply, including in a call transcript. You use it without leaving the app. If you use it, we save the reported reply, the 2 previous messages (yours included), the reason you choose, the language, the topic, the date and your account code. You can also report the example sentence on a flashcard or the translation in a pop-up by pressing and holding it: in that case we save the sentence, the word it refers to, the reason, the language, the date and your account code. We read the reports ourselves, to correct the AI's instructions and filters and remove wrong or offensive content. We don't send them to Anthropic.

2.7 Plan limits and usage counters

To apply the limits of your plan, we count on our server:

• chat messages per day and per month;
• voice conversation minutes per month, in real time and turn-based, including during the free trial;
• whether you have already used the Free plan's trial call;
• how many words you have saved;
• how many languages you study;
• for Listen and write, when each round starts and how many sentences it contains. The sentences you listen to and what you write stay on your phone;
• each day, the words looked up in the dictionary, the words and messages spoken aloud and how many new sentences you listen to, for the anti-abuse limits;
• each month, the characters read aloud and the voice clips recognised.

The limits of each plan are in the Terms (https://parlette.app/en/termini, section 6). The day is counted from midnight to midnight, UTC (in Italy from 1 a.m., or 2 a.m. during summer time), so we don't need your time zone.

We also count technical usage, meaning how much text the AI processes and how many voice minutes you use. This is for the anti-abuse limits described in the Terms and to keep costs under control.

The Free plan's trial call is one per account. To check this we don't use phone identifiers.

2.8 Subscriptions and purchases

Subscriptions are sold by Apple and Google: they collect payment, apply VAT and handle payments, refunds and cancellations. Through RevenueCat we receive:

• the plan, the duration (annual or monthly) and the start and end of the free trial;
• the renewal or expiry date and the status (active, cancelled, refunded);
• the product, the price and currency, the store country;
• the transaction identifier and the receipts.

The identifier we use with RevenueCat is your account's internal one, a code without name or email.

We never receive card or bank account details or your billing address: they stay with Apple and Google. We don't send RevenueCat your email, name or advertising identifiers.

On Android, if you buy with a creator's discount code, we send it to RevenueCat together with the purchase so the creator can be paid. If the purchase doesn't go through, we remove it.

If you withdraw from the contract with the "Withdraw from the contract here" command, we keep as proof of the withdrawal (art. 54-bis of the Italian Consumer Code) the name you write in the form, which may include your surname, the email for the receipt if it differs from your account's, the plan, the length, the store, the date the contract started and the date of the withdrawal. A copy of the receipt also goes to support@parlette.app. This data stays as long as you have the account.

We need this data to unlock your plan on all your devices, to restore purchases and to check that receipts are genuine. We also use it to see how subscriptions are doing, only as overall numbers.

We send you the notice before the annual renewal by email (point 2.1). The store may also send you its own, under its rules.

For accounting we keep the stores' sales reports. For each purchase they may show the order number, date, product, price, country, phone model and the discount code used, but not your name or your email.

If you haven't made any purchases, RevenueCat works out from your IP address the country you're in. It keeps only the country, not the IP, and we use it only in the overall subscription numbers.

2.9 Technical data

Every time the app talks to a server, that server sees your IP address. This happens with our server (including when you download pronunciation audio or the dictionary), with ElevenLabs during real-time voice conversations and with Cloudflare. We also process:

• operating system and version, app version, type of app or browser;
• your phone's language;
• our server's error logs.

This data is used to make the app work, to protect the service and to find bugs.

We don't collect the advertising identifier (IDFA, GAID) or other phone identifiers. On Android, ML Kit sends Google an installation identifier (point 2.5). We don't collect your precise location, contacts or photos; the country RevenueCat works out is in point 2.8. At launch we don't use crash reporting tools.

If you've allowed it in your phone's settings, Apple and Google collect statistics and crash reports on their own account and show them to us in aggregate form. They do this under their own privacy policies.

If you allow us, we use PostHog to understand which features are used. We ask you once, the first time you open Study, with two buttons: "Yes, that's fine" and "No, thanks". The app sends events such as "screen opened" or "challenge started", with a random code created on your phone that isn't linked to your account, your email or your name. The events don't contain what you write or say, we don't record the screen and the IP address is discarded. The data is stored in the EU, in Frankfurt, and we look at it only as overall numbers. Nothing starts before you say yes. You can change your mind in Profile > Usage stats: from that moment we stop sending events and delete the code from your phone.

2.10 Support and privacy requests

If you write to support@parlette.app we process your email, your name, the content of the message, attachments and our replies. For privacy requests we also keep what you asked and what we did. When you withdraw using the app's "Withdraw from the contract here" button, a copy of the request also goes to support@parlette.app.

The app opens your email app: the app itself doesn't collect anything.

Emails arrive in a Parlette inbox at Zoho (Zoho Mail, in the EU) and only the people who handle support read them. We reply from support@parlette.app.

2.11 The website parlette.app

The website doesn't use cookies, has no forms and doesn't use analytics. The fonts also come from the website itself, so the page doesn't load anything from other sites.

To show you the pages and protect them from attacks, Cloudflare processes your IP address, browser type, the page you request and the time. We don't keep website logs.

The website also hosts the Challenge link (parlette.app/sfida/…), which opens the app or takes you to the store, and the page to request account deletion (point 10).

2.12 What stays only on your phone

Some information stays only on your phone: the light or dark theme, the last conversation mode, the AI notices already shown, the daily streak, the answers to the first-launch questionnaire, scheduled notifications and your best score in Challenge practice. Before you sign in, your phone also keeps the languages you chose, the first word from Try it and the version of the Terms and of the AI line you accepted. They go into your account when you sign in.

After you sign in, your phone also keeps a copy of your account (languages, words, recent conversations) and the codes that keep you signed in, so the app works even with a weak connection. The dictionary for each language you download from the icon in Study also stays on your phone: it's the same for everyone and contains no data about you. Your phone also keeps the pronunciation audio of your words, up to 300 files, which is deleted when you sign out or delete your account, and the audio of the app's sentences, up to 1,500 files, which is the same for everyone and is deleted when you delete your account.

The copy of your account and the codes that keep you signed in are deleted when you sign out or delete your account. The rest stays on your phone until you uninstall the app. They are needed only to make the app you asked for work, so your consent isn't required (art. 122 of the Italian Privacy Code).

If you answered the usage stats sheet, your phone also keeps your answer (yes or no) with the date and, if you said yes, the random usage stats code (point 2.9).

The notification "Remember what … means?" shows a word you saved even on the lock screen, so people near you may see it.

If you are under 18 and the app has stopped, the phone remembers only the date of the block, for 30 days. It isn't sent to anyone.

2.13 Comments and messages on social media

If you comment "CODICE" under one of our Instagram posts, we automatically send you a private message with the discount code. To do that we see your Instagram username, the text of your comment and the message we send you. We use them only to send you the code you asked for. We don't send you other promotional messages unless you ask. We don't copy this data outside Instagram. On TikTok we don't send messages: the code is in the post caption. Comments are public and Meta handles them as a controller, under its own privacy policy. If you don't want to receive it, don't comment: you don't need the code to use Parlette.

3. What the processing is based on (legal bases)

The GDPR allows personal data to be processed only for one of the reasons listed in art. 6. For each purpose we use only one:

• contract: the data is needed to give you the service you asked for by accepting the Terms, or for steps you ask for before signing up (art. 6(1)(b));
• legitimate interest: we have a concrete interest, which you'll find written in the table, and we have weighed it against your rights (art. 6(1)(f));
• legal obligation: a law requires us to do it (art. 6(1)(c));
• consent: you said yes to us, and you can change your mind whenever you like (art. 6(1)(a)).

For every processing based on legitimate interest we have put in writing how we weighed it against your rights. If you're interested, you can ask us for a summary.

Only one processing is based on your consent: usage stats (point 2.9). Tapping "I agree" before the AI is the explicit permission required by Apple's rules: it tells you where your data goes and lets you choose. Under the GDPR we process that data on the basis of the contract. If you tap "Not now", you use Parlette with the AI off. Microphone and camera permissions are requested by your phone. They are needed to make those features work and you can remove them whenever you like.

4. What is required and what isn't

• Without an email, or without an Apple or Google account, we can't create your account.
• Without confirming that you are 18 the app can't be used.
• Without the AI permission you can't use chat, calls and Save word by voice. Grammar, saved words, flashcards and the Challenge remain.
• The microphone is optional: without the microphone you use Parlette in writing (point 2.4).
• The camera is optional: without it, you scan the Challenge QR with your phone's Camera app.
• Usage stats are optional: the app works the same if you say no.
• The subscription is optional: you can use the Free plan.

5. Artificial intelligence in Parlette

Where we use it

• In chat, in calls and in vocabulary building.
• In Practice and in Review with "My words".
• To transcribe what you say.
• To correct your messages and for pop-up translations.
• To recognise the language and base form of the word you save.
• For flashcard example sentences.
• To prepare the list of Challenge words.
• To score the written and spoken parts of the level test.
• For the voices of pronunciation and replies.

On screens with the AI, Save word included, you'll find the "AI" label. Every call starts by telling you that you're talking to an AI, and in Languages and voices it says that the voices are synthetic. Audio created by the AI carries a mark that computers can read, in the metadata. Written replies are short sentences and have no technical mark.

Who provides it

We build and offer Parlette's AI features ourselves: under the AI Act, Univapp is their provider. We use the models of two companies:

• Anthropic, with Claude, for text;
• ElevenLabs, for speech transcription, synthetic voices and live calls.

Before the app uses the AI for the first time, a line tells you that your written messages go to Anthropic (Claude) and your voice to ElevenLabs, including outside the EU, with links to "AI info" and to this policy. The details are in point 6. The line comes before the microphone request. The AI starts only when you tap "I agree": closing or going back doesn't count as acceptance. Anthropic receives your written messages and transcripts to produce the replies. ElevenLabs receives your voice to transcribe it, for live calls and for the synthetic voices. By contract both must protect your data at least as well as we do and as the GDPR requires (points 6 and 7).

What we send to the AI

• Your message and the transcript of your voice.
• Your native language, the language and level you're studying, the topic or the Practice rule.
• Your profile line (level, goal, interests, AI style).
• At most the last 8 messages of the same conversation, so the AI understands the context.
• With "My words", at most 20 words you have saved.
• The words you tap or type, when the dictionary doesn't have them yet.
• Your written answers and the transcript of the level test.

We don't send the AI your name, your email or your account identifier. To Anthropic and ElevenLabs we send only a code derived from your account, different from its identifier, which is used to detect abuse. After the account is deleted it no longer leads to you. In chat, voice notes, turn-based conversations and Save word, requests to Claude go through our server in Frankfurt, and the app doesn't communicate directly with Anthropic. In real-time conversations your phone connects to ElevenLabs, which also sees your IP address (point 2.4).

No training

We don't use your chats or your voice to train any AI, whether ours or the providers'. By contract, Anthropic can't use the content we send it to train its models. For ElevenLabs, we have turned off in our account the option that would allow it. In real-time calls, Claude works through ElevenLabs, which states in its contracts with model providers that they can't train on customer content.

So there is no training to object to. If you don't want your chats and your voice to reach the AI providers, turn off "Use AI" in Profile. To object to any other processing (art. 21 GDPR), write to support@parlette.app.

The providers' safety checks are a separate case: we explain them in point 6.

How long the providers keep data is in point 8.

Risks and precautions

• The AI can make mistakes: a correction or a translation may be wrong. Parlette doesn't replace a certified teacher, and levels A1 to B2 are not a certification. The AI doesn't give medical, legal or other advice and must not be used for emergencies.
• The AI may sometimes say something inappropriate or offensive. If that happens, use "Report".
• Text and voice also reach the United States, where authorities can access data under rules different from European ones. That's why we use the safeguards in point 7.
• Don't write or say sensitive data, such as information about your health, religion, political opinions or sex life. The same goes for passwords and other people's data. Conversations are for practice, so you can also make things up.
• If you happen to say or write sensitive data, it still reaches the AI providers together with the rest of the message. We don't use it for anything else and we don't save it. The providers keep it for the periods in point 8.

The AI has precise instructions: it talks only about the language you are studying, doesn't ask for personal data (name, age, address, school, health), avoids unsuitable content and treats what you write as an exercise, not as orders. The providers automatically block prohibited requests and replies, and a person at Univapp reads the reports.

Turning off the AI

You can turn off the AI whenever you like: in Profile, turn off "Use AI". From that moment the app no longer sends anything to the AI, and grammar, saved words, flashcards and the Challenge remain. Data already sent stays with the providers for the periods in point 8.

No automated decisions

The AI doesn't take decisions that have legal effects on you or that affect you in a similar way (art. 22 GDPR). You choose your level and you mark your own progress. In the Challenge, points are calculated by the server with fixed rules.

The level test is the only case in which an AI system gives you a score, and if you pass, your level goes up. It's an exercise: it produces no legal or similar effects on you (art. 22 GDPR), it doesn't issue certificates and nobody else uses it. If you think a score is wrong, write to support@parlette.app.

The details are in the AI Notice: https://parlette.app/en/informativa-ai

6. Who receives your data

Providers that work for us

These providers process data on our behalf and following our instructions (for Anthropic and ElevenLabs there is an exception for security, explained below the table). With each of them we have a contract (art. 28 GDPR) that obliges them to protect the data at least as well as we do. They may in turn use other providers (sub-processors), with the same obligations. If you want the list, ask us.

Anthropic and ElevenLabs process on our behalf the content we send them: texts, voice, transcripts. For one purpose only, security and abuse prevention, they decide for themselves how to use some data and are independent controllers, under their own privacy policies. When their checks flag content as contrary to their policies, they keep it for their own periods: Anthropic for up to 2 years, and the scores of these checks for up to 7 years; ElevenLabs for a period it doesn't publish (point 8). For ElevenLabs, the data needed to run the service is also its own, such as the IP address in real-time conversations.

Apple and Google

When you sign in with Apple or Google and when you buy a subscription, Apple and Google process your data as independent controllers: they decide how to use it, under their own privacy policies.

• Apple handles Sign in with Apple, the App Store, payment, refunds and cancellation. It also forwards our emails if you use "Hide My Email". In Europe your data is processed by Apple Distribution International Ltd (Ireland), which also sells the subscription. Privacy policy: https://www.apple.com/it/legal/privacy/
• Google handles Sign in with Google, Google Play, payment, refunds and cancellation. For people living in the EEA, Sign in with Google data is processed by Google Ireland Limited. On Google Play the subscription is sold by Google Commerce Limited. On Android, Google also receives the technical data from ML Kit, the QR code reader (point 2.5), and uses it under the ML Kit terms. Privacy policy: https://policies.google.com/privacy?hl=it

Meta (Instagram)

Meta Platforms Ireland Limited runs Instagram, comments and private messages, and handles them as an independent controller, under its own privacy policy: https://privacycenter.instagram.com/policy

Other Challenge players

In games with friends, the other players see your Profile name, your initial and your points. In Random they see your player name and your points. In your group's leaderboard, up to 30 people see your player name and your points. If you reach the top 100 of the league, all players in the league see them. They never see your email, your account or your Profile name.

Inside Univapp

Data is seen only by the people who work on Parlette and need it for their work. All of them are bound by confidentiality.

Advisers and authorities

If needed, we give the strictly necessary data to our accountant or our lawyers, who are bound by confidentiality. We give data to authorities only when the law requires us to.

We don't sell your data to anyone.

7. Data outside the European Union

Some providers are in the United States or access data from there. Zoho's technical support may access the support inbox from India. Every transfer is protected by at least one of these safeguards:

• the standard contractual clauses, a model contract approved by the European Commission (decision 2021/914) that obliges the provider to protect data as is done in the EU;
• the Data Privacy Framework, a decision of the European Commission (2023/1795) that recognises an adequate level of protection for participating US companies. We mention it only for providers that are listed as participants.

With all US providers we use the standard contractual clauses, including those that participate in the Data Privacy Framework. That way your data stays protected even if the Data Privacy Framework were to change. Each provider's safeguard is in the table in point 6.

We have assessed the risks of each transfer. You can ask us for a copy of the safeguards by writing to support@parlette.app.

8. How long we keep data

We keep data longer only if an authority asks us to or if it's needed in a dispute that is already open, and only for as long as necessary.

9. Your rights

You can ask us at any time, free of charge, for:

• access (art. 15): to know what data we have about you and get a copy;
• rectification (art. 16): to correct wrong or incomplete data. You can change much of it yourself in your Profile;
• erasure (art. 17): to have your data deleted. The quickest way is to delete your account (point 10), but you can also ask us to delete only some data, for example reports. The usage stats code isn't linked to you, so we can't find that data to delete it (art. 11 GDPR). You can, however, turn off the stats, and the code on your phone disappears;
• restriction (art. 18): to block the use of some data while we check a challenge you have raised;
• portability (art. 20): to receive the data you gave us in a JSON file, a common format that a computer can read, or to have it sent to another service when technically possible;
• objection (art. 21): see the box below;
• withdrawal of consent (art. 7(3)): if a processing is based on your consent, you can withdraw it whenever you like, as easily as you gave it. Today this applies to usage stats: you turn them off with one tap in Profile > Usage stats. What we did before remains valid.

You can remove phone permissions, such as microphone and camera, whenever you like from your phone's settings.

Your right to object

When we process your data for a legitimate interest of ours (the "Legitimate interest" rows in the table in point 3), you can object at any time on grounds relating to your particular situation. We stop processing it, unless we have legitimate grounds that override your interests and rights, or we need the data to defend ourselves in court. Just write to support@parlette.app.

How to exercise them

Write to support@parlette.app. If you can, write from your account email and tell us how you sign in (Apple, Google or email): that way we can tell straight away that it's you. We ask you for more information only if it's needed to verify your identity. If you use Apple's "Hide My Email", write to us from any address and tell us the Apple one (it ends with @privaterelay.appleid.com). We send a confirmation request to that address, and Apple forwards it to your real email.

We reply within one month. If the request is complex, or if we receive many, we may take up to three months in total. In that case we tell you within the first month, explaining why.

It's free. Only if a request is clearly unfounded or excessive, for example because you repeat it for no reason, may we ask you for a reasonable fee or refuse it, and we explain why.

Complaint

If you think we are processing your data improperly, you can lodge a complaint with the Garante per la protezione dei dati personali, the Italian data protection authority (www.garanteprivacy.it, Piazza Venezia 11, 00187 Rome). You can also contact the authority of the EU country where you live or work, or where the breach took place, and you can also go to court. If you like, write to us first and we'll try to sort it out together.

10. Deleting your account

In the app, go to Profile > Delete account > Delete.

If you no longer have the app, you'll find the steps at https://parlette.app/en/elimina-account. You can also write to us at support@parlette.app, if possible from your account email.

When you delete your account:

• we delete your profile, languages, progress, words, level test results, Challenge and league data, counters and sessions;
• on your phone we delete the copy of your account and the pronunciation audio of your words and of the app's sentences;
• we revoke Sign in with Apple, if you used it, and Sign in with Google if you delete your account from the app;
• if you have a subscription on Google Play, we cancel its renewal and then ask RevenueCat to delete your data. If we can't cancel it, the app asks you to cancel it yourself, and the data at RevenueCat stays until we have cancelled the renewal;
• in all other cases we immediately ask RevenueCat to delete your data;
• at ElevenLabs we delete the transcripts after each use; only technical and security logs remain, for the period in point 8;
• Anthropic deletes the texts by itself within 30 days, except those flagged by its safety checks (point 8).

We do it immediately, as soon as you confirm, and the app shows you "Account deleted". If you ask us by email, we reply when it's done.

Only what we must or may keep remains:

• the server's technical logs, which delete themselves after 1 day (1 hour for sign-in logs, point 8);
• reports about the AI, no longer linked to your account, until they expire. They may also contain your messages;
• a minimal note saying that the deletion took place, with the date and a code that no longer leads to you, so that we can prove it;
• emails to support, for the period in point 8, unless you ask us to delete them;
• the stores' sales reports, for the periods in point 8. For each purchase they may show the order number, date, product, price, country, phone model and the discount code used, but not your name or your email;
• your purchase history at Apple and Google, which they keep as independent controllers;
• texts flagged by Anthropic's safety checks, for the periods in point 8;
• usage stats, if you had said yes: they aren't linked to your account and are deleted when they expire (point 8).

Some information stays on your phone, for example the light or dark theme and your native language. It disappears when you uninstall the app.

On the App Store you must cancel yourself: deleting your account doesn't stop payments. Before you confirm, the app gives you the link to Apple's subscriptions page. Cancel at least 24 hours before renewal, otherwise the next period is charged.

Deleting your account doesn't mean withdrawing from the contract. In the first 14 days after subscribing, the app asks whether you also want to withdraw.

11. Minimum age

Parlette is for people aged 18 and over.

Before Try it, meaning before you use the AI for the first time, the app asks for your year of birth, which we don't save. By tapping "Continue" you declare that you are at least 18. The same applies to people who join the Challenge from the QR. If you are under 18, the app stops. The phone remembers only the date of the block: for 30 days you can't try again, even if you close the app, and support can't unblock you.

We don't analyse your conversations to work out how old you are.

If we find out that an account belongs to someone under 18, we delete it with all its data, without keeping any data about age, and we say so by email to the account address. If there was a subscription, on Google Play we cancel its renewal, on the App Store it must be cancelled in the store; a parent can ask Apple or Google for a refund. If you are a parent and think your child is using Parlette, write to us at support@parlette.app.

12. Security

• Data travels between the app and the servers over encrypted connections.
• In the database each account can read only its own data, thanks to row-level access rules.
• Provider keys are only on our server, never in the app. For calls, your phone receives an access code that expires shortly.
• Our database is also encrypted when stored on the servers (Supabase uses AES-256 encryption).
• The database and the technical logs stay in the EU.
• Data is seen only by the people who need it, and all of them are bound by confidentiality.

If a breach puts your data at risk, we notify the Garante within 72 hours. If the risk to you is high, we notify you too.

13. What we don't do

At launch, Parlette:

• doesn't show ads and doesn't track you, so the app doesn't ask you for Apple's tracking permission. PostHog usage stats start only if you say yes (point 2.9). The technical data Google receives from ML Kit on Android is in point 2.5;
• sends only notifications that your phone schedules by itself, for the daily streak and for reviewing words: they don't go through any server and you turn them off in your phone's settings;
• keeps league leaderboards for only 3 months.

14. Changes to this policy

When we change this policy we update the version and the date at the top. We tell you about important changes, such as a new provider or a new use of data, in the app 30 days before they take effect. If a change requires your consent, we ask you for it first. We send you previous versions on request: write to support@parlette.app.

You'll find this policy in the app (in Try it, at sign-in, on the Plans page and in Profile > Privacy) and on the website, at https://parlette.app/en/privacy. It is available in Italian, English, French, German and Spanish.