Privacy Policy

Mnema is operated by Yevhen Maltsev, Warsaw, Poland ("Mnema", "we"). We are the controller of the personal data described here. For any privacy question or request, write to hello@mnema.me.

Mnema is in a free test launch. Plans are given by invitation and nothing can be bought. The product is changing quickly, and this policy changes with it; the date at the top tells you when it last changed.

Mnema lets you build or import flashcards and quizzes, study them with spaced repetition or cram sessions, track your progress, share study material with other learners, compare your progress on a leaderboard, get study reminders, and ask Mnemo to create material from a topic, your own notes, documents and slides, photos of pages, or recordings of lectures. Each of these activities involves some of the data described below.

Account data. Your email address, a password (stored only as an Argon2 hash, never in clear text), your email verification status, and, if you turn it on, your two-factor authentication setup. The secret behind two-factor codes is stored encrypted; recovery codes are shown to you once and kept only in a form that cannot be read back. Each sign-in session records the IP address and the browser it was started from.

Your profile. A display name, an optional short bio and an optional profile picture. Other learners see them in the shared library, and your display name on the leaderboard. Until you choose a display name, they see a generated name instead, never your email address.

Sign in with Google. If you choose to sign in with Google, Google sends us your email address, whether Google has verified it, your name and your Google account's identifier, and we store them to create or find your Mnema account and to recognise you next time. On a new account your Google name becomes your display name, so other learners see it until you change it in your profile. We do not use your Google profile picture. We receive no access to your Gmail, contacts, files or anything else in your Google account, and we never post or read on your behalf. An account created this way has no Mnema password until you set one with the password reset link.

Study content. Decks, cards (text, formulas, code, pictures, sounds and image occlusion diagrams), quizzes and questions, tags, deck covers, and the files you import (Anki packages, CSV and text files). Also your review history: which cards you studied, how you rated them, the scheduling data derived from it such as intervals, streaks and retention, and your progress towards your daily goal. A review history that comes with an Anki package is kept with its cards, marked as imported, and never counts towards your streak, points or daily goal.

Material you give Mnemo. Documents (such as PDF, Word, PowerPoint, text and Markdown files), photos and other pictures, and audio recordings such as lectures. We store the file and the text read from it for as long as the chat that uses it exists. Recordings are transcribed into text, and pictures and scanned pages are read into text and a short description. When you approve cards or questions made from a PDF, a Word or PowerPoint file, a recording or a photo, that file is kept as their source, so study can show where a card came from: a Word or PowerPoint file as a PDF copy that our own converter makes (no outside provider sees it), a recording in a compressed copy, a photo resized and with its location and camera details removed. Kept files count towards your plan's storage and are listed in Settings → Billing → Sources.

AI interactions. Your chat messages, your answers to Mnemo's questions, the drafts it produces, your edits, approvals and undos, the web research a chat ran (the search queries and the text of the pages found, so a later request in the same chat can reuse them), and the credits each action used. For every AI card or question we store where it came from: the web page or your file, the page or the moment in a recording, and the passage it was built from.

Learner profile. Personalization is switched off during the test launch, and the app does not ask for a learner profile. A profile saved before it was switched off (native language, languages and levels, subjects, goals, preferred card shape, formatting and custom instructions) stays stored, but it is not sent to the AI and decides nothing.

Reminders and notifications. Which reminders you want and how (push, email or off for each kind), your daily goal, your time zone, and the hour you usually study, which we work out from when you studied over the last 14 days. If you allow notifications on a device, we store the delivery address your browser's push service created for that device and its encryption keys.

Credits. Your plan, your credit balance, and a record of every credit added or spent: what for, how many and when. If we invited you to the test, the email address the invitation went to, with the plan and credits it carries; and any bonus credits we gave you, for example as thanks for feedback.

Early access requests. If you ask for early access before you have an account, our request form collects your email address and, if you give them, your name, what describes you (for example student or working professional), what you study or work in, why you want to use Mnema, what you would mainly use it for, and how you heard about us, together with your agreement to this processing.

Bug and problem reports. When you use Report a bug: the description you write, up to two images you attach, and the platform and app version you were using. When you report a problem with an AI card or its source: the card or question it is about, the source page and the passage it was built from, the reason you picked, your note, and a copy of the AI chat that made it, as you saw it when you sent the report. Images you sent in that chat are kept with the copy; files you attached show by name only. We copy only your own chat, and a very long chat only in its most recent part. When we fix what you reported, we may send you a thank you note; to help write it, the text and images of the reports it answers may be sent to OpenAI.

Tester feedback. If you take part in the test, our feedback form asks for the email address you signed up with, which features you tried, your ratings and comments, what you would use Mnema for, what price feels fair, whether you would join a call, and your agreement. If you agree to be quoted, it also asks how to sign your quote (for example your first name and what you study), and we may email you to ask for a photo, which you can decline. If you book a call, we receive your name, email address and the time you picked, and we may take notes during the call. We do not record calls without asking you first.

Feedback in the app. Once you have used Mnema on a few different days, now and then, after you finish a study session or approve an AI draft, Mnema asks how it is going and offers our feedback form (see Tester feedback above), with bonus credits for filling it in. We store when we asked and whether you opened the form or chose Not now, so we wait a few days before asking again. You can also rate the cards or questions an AI draft made with a thumbs up or down and a short line. A rating keeps a copy of those cards or questions as they were when you rated them, so we can see what you rated even after you discard the draft or delete the chat. A thumbs down also keeps a copy of the AI chat that made the draft, as you saw it, with the images you sent in it, so we can see what went wrong. We copy only your own chat, and a very long chat only in its most recent part. Answering is always optional.

Team view. While Mnema is in its test, a member of the Mnema team may open your account in a read only view that shows your decks, cards, quizzes, review progress, AI chats and drafts the way you see them. We record who opened it, when, and which pages they looked at. The view never shows your password (we do not have it) or your two-factor secrets.

Shared library. Study material you choose to publish, with its name, tags and description; the ratings and reviews you leave on other items, shown with your display name and profile picture; reports you submit about other items; the moderation status of your own items; and a trust score built from your reports and published items, which decides how much weight your reports carry and whether publishing or reporting is paused for a while after abuse.

Leaderboard. Your study score and what it is made of (neurons, streak, reviews and the days you met your goal), shown to other signed-in learners next to your display name, this week and all time.

Billing data. Payments are switched off during the test launch: nothing can be bought, and no card details are collected. When payments open, Stripe will handle them; we will receive your subscription status and Stripe identifiers, never your card number, and Stripe will receive your email address.

Technical data. When you use Mnema, our hosting providers' systems and our servers log the IP address a request came from, the address requested, the time and the result, and errors with their technical details. We use these logs for security, debugging and abuse prevention. Errors are also reported to our error monitoring service (Sentry) without the content of your requests, cookies, passwords, tokens, email addresses or IP addresses. From the browser it receives the error, the page (without anything after the ? in its address) and which labelled buttons were pressed just before; never what you type, and it records no screen.

Browser storage. Signing in sets one essential cookie holding a session refresh token (HttpOnly, secure, same-site, valid for 30 days); signing in with Google sets a second one for 10 minutes to protect that sign-in. The short-lived access token lives only in memory. Your browser's local storage keeps conveniences on your device: theme, language, sidebar state, unsent chat text and answers, the chat or review you had open, your dictation language, audio and quiz timer settings, and the email address you are verifying. If you turn on notifications, a service worker is installed only to show them; it stores and caches nothing. We set no advertising or analytics cookies.

Dictation. Speaking to Mnemo uses your browser's built-in speech recognition. Mnema receives only the resulting text; the audio is handled by your browser and, depending on the browser, by its maker (Chrome, for example, sends it to Google).

• To provide the service you signed up for: storing and syncing your content, scheduling reviews, generating study material, keeping the sources of your cards, running the shared library and the leaderboard, and managing your plan and credits. Legal basis: performance of our contract with you.
• To keep the service secure and prevent abuse: logs, sign-in session records, rate limits and caps on AI use, moderation and spam detection in the shared library, and error monitoring. Legal basis: our legitimate interest in a safe and working service.
• To understand how the service is used: statistics such as sign-ups, visits, returning users, devices, generations run, credits and web searches used, retention and errors, viewed by the Mnema team in an internal dashboard. Some are kept per account, such as on which days you studied, how many cards you reviewed and how many drafts you made, so we can see whether Mnema helps people keep studying. Legal basis: legitimate interest. These statistics describe usage, not the content of your material, and are not shared with third parties.
• To send study reminders: at most one a day, only when there is something to do (cards due, a streak or daily goal at risk, drafts waiting for review). Legal basis: our legitimate interest in helping you keep the study habit you came for. They are on by default; you can turn each kind off in Settings → Notifications, and every reminder email has a link that stops reminder emails at once.
• To contact you about your account: verification and password reset emails, security notices such as two-factor changes, invitations to the test, bonus credit notices, billing receipts once payments open, and, during the test launch, messages about the test itself such as changes that affect your data. Legal basis: contract and legitimate interest. We do not send marketing email without your consent.
• To run early access: deciding whom to invite while Mnema is invite only, and emailing you your invite and news about early access. Legal basis: your consent, given on the request form, which you can withdraw at any time.
• To fix problems and improve Mnema: reading bug reports, problem reports and the chat copies attached to them, feedback answers and call notes, replying to you by email, and adding the bonus credits we promise for feedback. Legal basis: our legitimate interest in a working service for bug and problem reports; for the feedback form and calls, your consent, given on the form or by booking a call, which you can withdraw at any time; for feedback in the app, our legitimate interest in improving Mnema, and answering is always optional.
• To help you and to improve Mnema through the team view: while Mnema is in its test, the Mnema team may open your account read only, for example to answer a question you sent us, to look into a problem, or to understand how testers use Mnema and where the AI falls short. The view cannot change anything in your account, send messages, spend your credits or sign you out, and it does not count as you using Mnema. Only named members of the team can open it, and every opening is logged. Legal basis: our legitimate interest in supporting you and improving a product that is still being tested. You can object at any time by switching Team view off in Settings → Security, or by writing to us; we then stop viewing your account.
• To quote you: showing your quote on our website, signed the way you chose, with your photo only if you agreed to that too. Legal basis: your consent. If you withdraw it, we take the quote down.
• To comply with the law, for example tax records for purchases once payments open. Legal basis: legal obligation.

When you use Mnemo, these are sent to OpenAI's API: what you type, the recent part of the chat and a summary of the earlier part, the text of the files you attach, your pictures and scanned pages (to be read), your recordings (to be transcribed), the names of your decks and quizzes, and up to 50 existing cards of the deck being filled, so new cards do not repeat them. Material you publish to the shared library, and text read from your files and pictures, is also checked by OpenAI's moderation service. Your learner profile is not sent while personalization is switched off.

When you ask about a topic without giving material, search queries written from your request are sent to a web search provider (Tavily). It receives the queries only, never your files, your chat or who you are, and returns the text of public web pages, which Mnemo reads and cites on each card. For medicine and health requests, a few words from the subject also go to Europe PMC (open access medical literature) and openFDA (United States drug labels). When a diagram would help, the topic goes to Openverse, an open license image search, and our server downloads the image it finds from the site that hosts it (for example Wikimedia Commons); we attach only permissively licensed images and show their attribution. A word on a vocabulary card may be looked up on Wiktionary to check its translation. None of these receive your name, your email address or your files.

OpenAI's API data policy says data sent through its API is not used to train its models unless the customer opts in to share it, and that it keeps API data for up to 30 days, for abuse monitoring and to continue a conversation, before deleting it unless the law requires longer. Tavily's privacy policy allows it to use query data to improve its own search results. We never use your content to train models.

Generated material is stored as drafts so you can review, edit, approve or discard it. We keep your 30 most recent generations and clear older ones; a draft you never reviewed stays until it is 30 days old.

AI output can be wrong, and a cited source may not say what a card claims. It is study material for you to check, not medical, legal or professional advice. Please review generated cards and questions before relying on them.

We share personal data only with providers that process it for us, and only what each one needs:

• OpenAI (United States): AI models for chat, writing cards and questions, reading pictures and scanned pages, transcribing recordings, moderation, and helping us write thank you notes for fixed bug reports.
• Tavily (United States): web search for research-based generation, receiving search queries only.
• Europe PMC (run by EMBL's European Bioinformatics Institute) and openFDA (the United States Food and Drug Administration): sources for medicine and health requests, receiving a few search words only.
• Openverse and the sites that host the images it finds, and Wiktionary: open-licensed diagrams and translation checks, receiving a topic or a single word only.
• Heroku, a Salesforce company: hosting of our servers, background jobs and database, with the database's backups.
• Cloudflare: hosting of the website (Cloudflare Workers), storage of the files, pictures and sounds you upload (Cloudflare R2), our domain, and receiving the email you send to a mnema.me address (Email Routing).
• Resend: delivery of our emails, including reminders.
• Sentry (Functional Software, United States): error monitoring, as described in section 3.
• Your browser's push service (Google, Mozilla, Apple or Microsoft, depending on your browser): delivering study reminder notifications. Each notification is encrypted for your browser.
• Google (Sign in with Google): confirming who you are when you choose to sign in with your Google account. Google receives the fact that you signed in to Mnema; we receive your email address, name and account identifier.
• Google (Google Forms and Google Sheets): collecting and storing early access requests and tester feedback.
• Google (Google Calendar and Google Meet): booking and holding feedback calls, if you choose to book one.
• Our email provider: answering the email you send us.
• Stripe: subscriptions, credit packs and invoices, once payments open. Nothing is sent to Stripe while payments are switched off.

Other learners see what you make public: items you publish to the shared library, with your display name, bio, profile picture, leaderboard rank and number of published items, under the license described in the Terms of Service; the reviews you leave there; and your display name and study score on the leaderboard. The shared library and the leaderboard are visible only to signed-in Mnema users. We may also disclose data when the law requires it, or to protect the safety of users or the service. We do not sell personal data.

Several providers above are based in the United States or operate globally. Where personal data leaves the European Economic Area, we rely on the safeguards the law provides, such as the EU-US Data Privacy Framework or standard contractual clauses, and on the provider's security commitments.

• Account, profile and study content: for as long as your account exists.
• Sign-in session records (IP address and browser): for as long as your account exists.
• Files you give Mnemo: an upload that never reaches a chat is deleted within about 24 hours (an upload that never finishes, within about 2 hours). A file in a chat stays as long as the chat. A file kept as a source stays while at least one card or question made from it exists, or until you remove it in Settings → Billing → Sources, and is then deleted within about an hour, unless a chat you still have is using it.
• Pictures and sounds you remove from your cards: deleted within about an hour.
• AI chats: kept while you use them. A chat you have not opened or written in for 90 days is removed automatically, unless one of your 30 most recent generations came from it. A chat you delete is erased at once, with its history and any draft still waiting for review, and the files only it used follow within about an hour; the cards and quizzes you saved from it are not affected. Either way, if you reported a problem with something the chat made, or gave its draft a thumbs down, the copy kept with that report or rating stays until it is deleted.
• AI drafts and the web research saved with them: your 30 most recent generations are kept; older ones are cleared, except that a draft you never reviewed stays until it is 30 days old. The cards and quizzes you approved are study content and stay.
• Exports: the file is deleted 24 hours after it is made, and its download link works for one hour.
• Early access requests: until early access ends, then deleted; sooner if you ask us or withdraw your consent.
• Bug and problem reports, with any chat copy attached: for as long as your account exists, so we can follow up; erased with your account, or sooner if you ask us. The images attached to a bug report are deleted 30 days after the report is closed as fixed.
• Tester feedback and call notes: until 12 months after the test ends, then deleted; sooner if you ask us or withdraw your consent.
• Feedback in the app (when we asked you, your thumbs and lines on AI drafts, the copy of the cards or questions a rating keeps, and the chat copy a thumbs down keeps): for as long as your account exists; erased with your account, or sooner if you ask us.
• Records of team views (who opened your account, when, and which pages): 12 months, then deleted; erased with your account.
• Quotes and photos: for as long as we show them. If you withdraw your consent, we take them down and delete the photo.
• Notification addresses for your devices: until you turn notifications off on that device, the browser withdraws them, or your account is erased.
• Server logs and error reports: for a limited period needed for security and debugging.
• Billing records: for as long as tax and accounting law requires.
• After you delete your account: it is deactivated at once and erased 14 days later (logging in within those days restores it). Erasing deletes your study content, files, chats, drafts, reports, feedback, notification settings and sign-in records, and removes your email address and password. What stays: items you published to the shared library, under your display name, with the reviews and reports you left there (unpublish them, or ask us, first if you want them gone), and your credit records. Copies in backups follow their normal rotation.

Under the GDPR and similar laws you can ask us for access to your data, correction, deletion, restriction and portability, and you can object to processing based on legitimate interest. You can withdraw any consent you gave at any time.

• Export: download your decks and quizzes as an Anki package, CSV or text files from Settings → Data at any time, with your review progress in the Anki package if you tick it. For a copy of anything else we hold about you, write to hello@mnema.me.
• Sources: see the files kept for your cards in Settings → Billing → Sources, and remove any of them. Their cards keep their text and passages.
• Chats: delete any AI chat from the recent chats list. It is erased at once, with its history and any draft still waiting for review; the cards and quizzes you saved from it are not affected. A copy attached to a problem report you sent, or kept by a thumbs down you gave, stays with that report or rating until it is deleted; ask us and we delete it.
• Team view: see in Settings → Security when the Mnema team last viewed your account, and switch Team view off there at any time.
• Profile: change your display name, bio and profile picture from your profile in the sidebar at any time. Other learners see the change straight away.
• Reminders: choose push, email or off for each kind in Settings → Notifications, or stop reminder emails with the link in any of them. Notifications on a device can also be blocked in that browser's settings.
• Preferences: edit language, appearance and other settings in Settings.
• Feedback and quotes: to have your feedback answers deleted, or a quote or photo taken down, write to hello@mnema.me. Nothing else in your account changes.
• Deletion: delete your account in Settings → Account. It is deactivated immediately and erased after 14 days; logging in within those days restores it. Items you published to the shared library stay available to others; unpublish them before you delete your account if you want them gone.
• Complaints: you can also contact your data protection authority.

To make a request, write to hello@mnema.me. We may ask you to confirm your identity through the email address on the account.

Data travels over encrypted connections. Passwords are hashed with Argon2. Sign in with Google uses the standard OAuth flow with PKCE, and we accept a Google account only when Google confirms its email address is verified. Access tokens are held in memory only and refresh tokens in an HttpOnly cookie, stored on our side only as a hash. Two-factor authentication is available, with encrypted secrets and one-time recovery codes. Source files you keep open only through signed links that expire after 15 minutes. Pictures and sounds in cards, quizzes and deck covers, and profile pictures, are served from long, unguessable addresses: anyone who has the exact address can open the file, so keep sensitive images out of your cards. Access inside the team is limited to what operating and improving the service requires. The read only team view is open only to named members of the team, and every use of it is logged. No system is perfectly secure; if a breach affects you, we will inform you as the law requires.

Mnema is not directed at children under 16. We do not knowingly collect their data, and we delete accounts we learn belong to a child.

We will update this policy as the product evolves, especially during the test launch. Material changes are announced in the app or by email before they take effect, and the date at the top always shows the current version.

Yevhen Maltsev, Warsaw, Poland. To reach us, write to hello@mnema.me.