Privacy policy
Last updated 24 September 2026.
Fuego is an AI language tutor. To teach you, it has to store what you say to it. This page states exactly what that means: every category of data we keep, every company that sees it, where it sits, how long it stays, and how to make us delete it.
Who we are
The data controller for Fuego is:
Filip Kowalski
c/o Coworking Minds
Carrer Ferrers, 1
07510 Sineu
Islas Baleares, Spain
Questions about anything below go to privacy@heyfuego.com.
What we collect
Account
The app does not require registration. On first launch it creates an anonymous Firebase Authentication session, and we store the Firebase user identifier assigned to it. That identifier is the sole means by which your learning data is associated with you between sessions. We also store the name you provide during onboarding. Anonymous sign-in does not require an email address or password.
Because the identifier is held on the device and nowhere else, deleting the app, or deleting your account through the app, severs the association between you and the data described below.
Learning profile
Your native language, the language you are learning, your time zone, your self-reported level, and the level the tutor assesses from your conversations. The tutor also builds a memory of you from your conversations: your goals, your interests, your recurring mistakes, and preferences you state out loud. We keep a version history of that memory. You can read and correct it inside the app.
Conversations
Every chat message you send and every reply the tutor gives, stored as text. This includes voice sessions: the spoken conversation is transcribed and the transcript is saved as text in the same place as your typed chats.
Practice
Your flashcards and their review schedule, the exercises generated for you, the answers you submit, and the proficiency estimates we derive per skill over time.
Device and billing
Where you permit notifications, we store a push token and the platform it was issued by.
Fuego offers limited access at no charge and optional auto-renewing subscriptions sold through the App Store and Google Play. When you buy or restore a subscription, we store the plan, entitlement status, renewal dates, and store transaction identifiers needed to grant access and prevent duplicate grants. Apple or Google processes the payment itself; we never receive your card number. We record voice-session counts and duration to understand feature use.
Usage analytics and attribution
We record app opens, onboarding progress and structured choices (languages, starting level, learning goal, daily practice minutes and notification preference), feature starts and completions, conversation and message counts, voice-session duration, and purchase-flow events. These metrics are associated with your app identifier and help us understand how Fuego is used. Our analytics events do not contain conversation or feedback text. Each analytics event also carries technical information about the device on which the app runs: the platform, the app version and build number, the operating system version, the device model (for example, "iPhone17,3"), the device language, the region setting and the timezone. The device model identifies a product line and does not identify an individual device. We send your supplied name, onboarding completion and choices, interface locale, timezone, preferred practice time and aggregate usage counts to RevenueCat for customer segmentation. We also use a Firebase Analytics app-instance identifier, supported Google Play install-referrer fields and Apple ad-attribution information to associate subscriptions with acquisition campaigns.
Feedback
If you send feedback from inside the app, we receive your message, the email address you type in, your platform, app version, and locale.
Voice and audio
We do not store your recordings. Audio is streamed live and then gone. What survives is the text transcript.
When you talk to the tutor, your audio streams to our own voice relay server in Germany, which then passes it to our processors: Soniox for speech recognition, Google for the tutor's reply, and Soniox again for the spoken answer. The relay holds no database and writes no files. Audio passes through memory and is discarded.
Only text comes back to us: the transcript lines and the number of seconds you spoke, which we use to track your progress.
Audio the tutor speaks in listening exercises is synthesised, cached in Cloudflare R2, and shared between learners who get the same sentence. That cache holds no recording of you.
How we use it
- To run the tutor: your profile and conversation history become the context we send to the language model.
- To generate lessons, exercises, and flashcards pitched at your level and your weak spots.
- To track your streak and your progress, including how much you speak.
- To send push notifications you asked for.
- To answer your support messages.
- To find crashes and fix them.
We do not sell your data. We do not share it with advertisers. We do not use your conversations to train our own models.
Companies that process data for us
Each of these sees part of your data to do a job for us. None of them may use it for their own purposes.
| Company | What it receives | Why |
|---|---|---|
| Google (Firebase) | Your email, sign-in identifiers, push token, crash reports tagged with your user ID | Sign-in, notifications, crash reporting |
| Google (Gemini) | Your messages, your learning profile, your tutor memory, and your voice session's recognized text | Generating the tutor's replies and exercises |
| Soniox | Your live audio, via our voice relay, and the tutor's reply text | Speech recognition, and speaking the tutor's replies and listening exercises aloud |
| RevenueCat | Your user ID, purchase records, supplied name, onboarding completion and structured learning preferences, locale, timezone, aggregate usage counts and app-install attribution identifiers | Managing subscriptions, customer segmentation and purchase attribution |
| Cloudflare | Requests to this website; cached synthesised lesson audio | Serving the site and lesson audio |
| Hetzner | Everything in our database, as the operator of the server it runs on | Hosting |
| Telegram | Feedback you send us, and transcripts of voice sessions | Alerting us to problems and reviewing tutor quality |
Apple or Google processes subscription payments and RevenueCat manages the resulting entitlement. They receive the identifiers and purchase records needed to complete, restore, renew, or refund a subscription. We do not receive your card number.
Where your data is stored
Our database and our voice relay run on a Hetzner server in Falkenstein, Germany. Backups are copied to Hetzner Storage Box, Cloudflare R2, and Backblaze B2. Our processors are global companies, so your data does reach servers outside the European Economic Area, in particular in the United States.
How long we keep it
Be clear on this one. Nothing expires on a schedule today. Your conversations, transcripts, flashcards, and tutor memory stay until you ask us to delete them or you delete your account.
Backups are kept for 7 days, 4 weeks, and 3 months on a rolling basis. So for up to roughly three months after a deletion, a copy can still exist inside a backup we cannot edit. Those copies are not used to serve the app and roll off on their own.
Your rights
If you are in the UK, the EEA, or a US state with a privacy law, you can ask us to show you your data, correct it, export it, delete it, or stop a particular use of it. You can complain to your national data protection authority.
Email privacy@heyfuego.com from the address on your account. We answer within 30 days. There is no charge.
For deletion, follow the steps on the account deletion page. It says what gets removed and what does not.
Our legal basis, if you are in the EEA or the UK
- Contract. Running the tutor, storing your progress, and managing your subscription. Without this data there is no product.
- Consent. Push notifications, and the microphone. You grant both at the OS level and can withdraw either in your phone's settings.
- Legitimate interests. Crash reporting, abuse prevention, and reviewing tutor output for quality.
Children
Fuego is not built for children. Do not use it if you are under 13, or under the minimum age of consent where you live if that age is higher. We do not knowingly collect data from children. If you believe a child has an account, write to privacy@heyfuego.com and we will delete it.
Security
Traffic runs over TLS. Sign-in is handled by Firebase, so we never hold a password. Access to the production database is restricted to the operator over a private network, and the server does not accept public SSH.
No system is perfect. If you find a security problem, email security@heyfuego.com and give us a way to reach you.
Changes
When this policy changes we update the date at the top. If a change means we start doing something materially different with your data, we will tell you in the app before it takes effect.
Contact
privacy@heyfuego.com for anything on this page. support@heyfuego.com for everything else.
By post: Filip Kowalski, c/o Coworking Minds, Carrer Ferrers, 1, 07510 Sineu, Islas Baleares, Spain.
Our lead supervisory authority is the Agencia Española de Protección de Datos, www.aepd.es. You may complain to it, or to the authority in your own country.