Privacy Policy

Last updated 1 August 2026

Nami is an AI assistant that talks to you in the language you’re learning. To do that it has to read what you write, listen to what you say, and remember what you saved. This page says exactly what that means, who else sees it, and how to erase it.

Nami is built and run by Marcin Zurek, an independent developer. For anything on this page, write to mz.mzurek@gmail.com.

What Nami collects

Your account

Signing up requires an email address and a password. Those are held by Supabase Auth, not by Nami’s own server — Nami’s server only ever receives an anonymous user ID from your login token, and never sees your email address. Email and password are the only sign-in method; there is no Sign in with Apple or Google.

What you create in the app

• Chat messages — everything you type or dictate, and Nami’s replies, with the per-word breakdown used by the magnifier.
• Voice recordings — the audio file of every voice note you send and every dictation you make, plus its transcript.
• Photos you attach to a message — the picture itself, kept with the chat it belongs to. Your phone shrinks it before it is uploaded, and location data (EXIF) is stripped on the device: the photo’s GPS coordinates never leave your phone.
• Saved words and expressions, their meanings, and your review history and schedule.
• To-dos and reminders you ask Nami to keep.
• Your profile — your level, the languages you picked, and anything you write in the free-text “about you” box.
• Notes Nami takes about you. To be a useful assistant, Nami extracts short facts from your own messages — an upcoming trip, a person you mention, something you’re into — and keeps them as memory notes. Notes about dated events expire automatically seven days after the event. You can read and delete these notes in the app at any time, individually or all at once.
• Feedback on your language — for each message you write, a stored analysis with the mistakes found, the corrections, and a score; plus practice exercises generated from your saved words, a short daily written summary of how you’re doing, and which words you’ve used and how often.
• Usage counters — your streak, daily challenge and story records, how many times you opened the app on a given day, and a running tally of how much your AI usage cost (used to enforce a per-user daily budget).

Device permissions Nami asks for

• Microphone — only while you are recording a voice note or dictating.
• Camera — only when you tap the + button and choose to take a photo. Nami cannot open the camera on its own.
• Photos — when you attach a picture from your library, iOS shows you its own picker and hands Nami only the photo you picked. The app never gets access to your library and cannot see any other picture in it.
• Notifications — so Nami can remind you to practise. Every reminder is scheduled on your own phone; no device token is registered anywhere and the server never knows what your phone showed you or when.
• Motion sensors — so Nami’s sun leans her gaze toward how you hold the phone. This never leaves your device.

Nami does not ask for location, contacts, calendar or files, and does not use Apple’s advertising identifier.

Who else sees it

Nami is a small app standing on other people’s infrastructure. These are all of them.

Your messages are not used to train AI models. Nami’s OpenRouter account has training turned off, which means requests are never routed to a provider that trains on what you send. OpenRouter itself does not store your prompts or replies — the settings that would allow that are off — though it does keep metadata about each request (token counts, latency, cost), never the content. Providers may hold a request briefly for abuse monitoring under their own terms, which is standard across the industry. Still, the sensible rule with any AI assistant: don’t tell it something you would not want a third party to hold.

Analytics: what actually leaves the app

Nami uses PostHog to count events — “a message was sent”, “a word was looked up”, “onboarding finished”. What matters is what rides along with those counts, and the answer is: almost nothing.

• Properties are sent only if their name is on a hard allowlist, so no part of the app can leak content by accident. It holds things like which languages you picked and whether an answer was correct.
• No message text, no vocabulary, no to-do text, and no email address is ever sent. Message length goes as a bucket (short / medium / long), never the message.
• You are identified by your anonymous account ID (a random UUID), plus your OS version and time zone. Nothing is tracked before you have an account.
• Session replay is off. Autocapture is off. Nami does not record your screen, your taps, or your keystrokes.

Nami’s own server additionally records the days you opened the app, to measure whether people come back.

Crash reporting

Sentry receives crash and error reports from the app and the server: stack traces, the component that failed, and technical breadcrumbs. Personally identifying information is explicitly disabled on the app side. Server-side error reports may be tagged with your anonymous account ID. One narrow case can include content: when a generated practice sentence fails Nami’s internal text checks, up to 60 characters of that AI-written sentence — built around a word you saved — is included in the report. Your own writing is never sent to Sentry.

Your photos

When you attach a picture, your phone shrinks it and strips its location data first, then uploads it to Nami’s server, which stores it in a private Supabase bucket and sends it to the AI model so Nami can answer you about it. It is served back only to you, and only with your login token — the file has no public address, and nobody else can fetch it.

A photo is deleted with the chat it belongs to, or with your account. There is no automatic expiry. Nami sends a photo to the model on the message you attached it to and, so it can answer a follow-up question about the same picture, on the turn after that — then it stops being part of the conversation sent to the model, though it stays in your chat.

Your voice recordings

When you send a voice note or dictate, the audio is uploaded to Nami’s server, sent for transcription (via OpenRouter), and then saved as a file on the server. It is served back only to you, and only with your login token — nobody else can fetch it.

Voice-note audio is deleted with the chat it belongs to, or with your account. Dictation audio is kept until you delete your account; those audio-plus-transcript pairs are used to improve transcription quality. There is no automatic expiry.

How long things are kept

• Everything in your account: until you delete it, or delete your account. There is no automatic clean-up, with one exception: memory notes about dated events disappear seven days after the event.
• Server logs (timings, error traces, anonymous account IDs) are kept by the hosting provider on its own schedule.
• Analytics and crash reports live in PostHog and Sentry under their own retention schedules. Deleting your account does not automatically erase them — email me and I will purge them by hand.

Deleting your account and everything in it

Open Settings → Delete account in the app and confirm. This is not a request form; it runs immediately and it is permanent:

• your login is deleted from Supabase;
• every row belonging to you is deleted from the database — chats, messages, assessments, saved words, exercises, to-dos, memory notes, tutor notes, profile, streaks, challenges, stories, word history, usage counters;
• every audio file of yours is deleted from the server’s disk, and every photo you attached is deleted from storage.

There is no tombstone, no grace period and no way back. If the server cannot finish, nothing is deleted and the app tells you to try again — it never half-deletes. What it cannot reach is analytics and crash data (see above) and any copies held by the AI providers under their own policies. You can also delete individual things without deleting your account: a chat, a saved word, or any single memory note Nami has taken about you.

What Nami does not do

• No advertising, no ad networks, no attribution SDKs, no tracking across apps.
• No selling or renting of personal data. Ever, under any circumstances.
• No push notifications from a server — reminders are local.
• No payment details ever reach us. Nami sells a subscription through the App Store, so Apple takes the payment and we are told only that a subscription exists and when it ends — never a card number, a billing address, or a name.
• No screen recording, no keystroke capture, no session replay.

Your rights

If you are in the EU/EEA or the UK, you have the right to access, correct, export or erase your personal data, and to object to or restrict its processing. Deletion is available to everyone in the app, in one tap. For anything else, email mz.mzurek@gmail.com and you will get a reply from a person. Data is stored in the EU, but the AI providers, analytics and crash reporting named above may process it outside the EU, including in the United States.

Children

Nami is not intended for children under 13 (or under 16 in the EEA), and accounts are not knowingly created for them.

This website

The page you are reading is static. It sets no cookies, runs no analytics, and loads nothing from a third party.

Changes

If what the app does changes, this page changes with it and the date at the top moves. Nami is in active development and early beta, so expect that to happen.