Privacy policy
Effective date: 25 October 2026
Fusen is made by lapis, an individual developer ("we", "us"), who is responsible for your personal information. This policy explains what information the Fusen app for Android, its desktop companion app, the Fusen account service and this website collect, why, and what you can do about it.
Questions or requests: support@fusen.ink.
The short version
- Your notebook (saved words, sentences, notes, collections, the terms you add to them, and images saved with words) stays on your device. If you turn on sync, it goes to a hidden folder in your own Google Drive; images go there only if you also turn on Sync images. We never receive it.
- The dictionary is downloaded once and searched on your device. Looking up a word sends nothing anywhere.
- If you sign in to an account, we store your email address, your credits and records of how many credits each request used.
- When you use Fusen AI, the passage you translate or explain, or the image you ask it to read, is sent to our server and to the AI provider whose model answers it, used to answer you, and not kept by us. The AI providers we use are listed below.
- Two things you choose to send are kept, for 180 days: an answer you rate (thumbs up or down), with the text it was about, and a problem you report, with the lookup, the text and any images it includes. We use them only to improve Fusen.
- If you buy credits or a monthly plan, Stripe handles the payment. We never see your card number.
- The app sends anonymous daily usage counts with no identifier and no text. You can turn this off.
- We do not sell your information, show ads or use your information for advertising.
Information that stays on your device
The following is stored only on your device and never sent to us:
- Your notebook: saved words, the sentences and translations saved with them, notes, collections, and each collection's work title and terms.
- Images saved with words (a page you read text from, or a photo or screenshot you added). Android's own backup to Google leaves them out; notebook backups include them, and sync sends them to your Google Drive only if you turn on Sync images. Asking Fusen AI to read the text in an image, or attaching it to a report, sends it as described below.
- The dictionary, its index and any definition packs you install (for example the Chinese definitions).
- Your settings.
- API keys you enter for DeepL, Anthropic or an OpenAI-compatible server, and your account's session token. These are encrypted on the phone and left out of Android's backups and of notebook backups.
Notebook backups you export are files you control. We do not receive them.
Information we collect
When you sign in to an account
Signing in is optional. You need an account only for Fusen AI: translation, explanations and reading text from images. You can sign in with Google or, where the app offers it, with an email address.
- Account: your email address (from Google, or the address you entered), your Google account ID if you signed in with Google, when the account was created and last used, and its status.
- Sign-in sessions: for each signed-in device, a session record with the device's name (on Android, the phone model), its platform, and when the session was created and last used. A session that is not used for 180 days expires.
- Credits: your free and paid credit balances and the credits used this month.
- Usage records: for each AI request, the time, the kind of request (translation, explanation or text recognition), the model, the target language, the kind of text (for example a novel), the length of the text, the amount of AI processing used, its cost, the credits charged, whether the AI declined the text, and the app version. For text recognition, also the image's size (in bytes and in pixels), the length of the text read and the number of speech bubbles found. The text and the image themselves are not recorded. We use these records for billing, support, preventing abuse and setting prices.
- Email sign-in: to send you a sign-in code, we store your address and a hashed copy of the code until it expires. The code is sent by our email service provider (see Service providers), which keeps addresses that bounced or reported our mail as spam on its suppression list. To stop the sign-in from being used to send spam, we count requests per network using a hashed form of your IP address that changes every day and is deleted after 2 days. We do not store your IP address.
When you use Fusen AI
When you translate a passage or ask for an explanation with Fusen AI, the app sends our server:
- the passage (at most 2,000 characters);
- for context, up to five passages you looked up just before it in the same collection;
- for explanations, the word, its reading and the dictionary's meanings for it;
- for translations, the collection's work title and the terms you added to it (such as character names and places), so that names come out the same on every page;
- the target language and the kind of text (the collection's kind, for example "novel").
When you ask Fusen AI to read the text in an image (for example a manga page, a screenshot or a photo), the app sends our server the page or screen you chose, as a JPEG image, and, if you tapped a spot on it, where you tapped. For a tap, our server draws a ring there on a copy of the image so that the AI knows which text you mean. The copy sent to the AI carries none of the image's metadata (such as where a photo was taken).
This happens only when you ask for it, and only while you are signed in to a Fusen account. Our server sends the request to the AI provider whose model handles it, returns the answer to you, and does not store the text or the image. Our server logs record the request (time, route, result, processing used and cost, and for an image its size) but never the text or the image. Logs are kept for 30 days.
AI providers. Fusen AI uses models from the AI providers in the table below, through their APIs. Which provider answers a request depends on the model Fusen AI uses for it. The provider receives the text and images described above, never your email address or anything else that identifies you. What each provider may do with them, such as how long it keeps them and whether it may use them to improve its models, is in the table and in that provider's own policy.
| AI provider | Where | Use and retention |
|---|---|---|
| Anthropic, PBC | United States | Does not use data sent through its API to train its models. Keeps it only for a limited time, longer if it is flagged under Anthropic's usage policies. See Anthropic's privacy policy. |
Before Fusen AI sends your text or images to an AI provider that is not in this table, we add it here, and we tell you in the app before the change applies.
Do not send text or images you would not want processed by an AI provider. A screenshot shows everything that was on the screen, such as notifications and the time. Choosing a different translation engine, or turning off Translate by default, means nothing is sent until you tap the translate button.
When you rate an answer
Where the app offers it, you can rate an answer from Fusen AI with a thumbs up or down and, after a thumbs down, say what was wrong. The app then sends our server, linked to your account:
- the kind of answer (translation, explanation or a page read from an image), your rating, the reason you picked and any note you wrote;
- the answer and what it was about: the passage, the word, the AI's text and, for a page read from an image, each speech bubble's text and translation (not the image itself);
- the translation engine, the model, the target language, the app version, and your phone's model and Android version.
We use ratings only to find and fix bad answers and to compare models and prompts. A rating is kept for 180 days, then deleted, and is deleted sooner if you delete your account.
When you report a problem
Where the app offers it, you can report a problem with a lookup. A report sends our server, linked to your account:
- what was on screen: the passage (at most 2,000 characters), the dictionary entries shown, the translation and the explanation;
- where the lookup came from: the reading app, the collection's kind, and whether it came from an image;
- the description you write;
- the app version and language, and your phone's model and Android version;
- the images you choose to include, such as the page you read text from.
Images are kept in private storage that only the developer can read. We use reports only to fix the problem reported and similar ones. A report is kept for 180 days and its images a day longer, then deleted; both are deleted sooner if you delete your account.
When you buy credits or a monthly plan
You can buy credit packs and monthly plans in the app. You pay on Stripe's checkout page, which the app opens in a browser tab. Stripe (Stripe, Inc. and its affiliates) processes the payment as our payment processor: your card or other payment details go to Stripe, never to us. Stripe may ask for a billing address. Stripe's own privacy policy covers what it collects.
- What we send Stripe: the product, its price and an internal ID that links the payment to your account. For a credit pack, also your account's email address if it has one, so that Stripe can send the receipt.
- Monthly plans: Stripe keeps a customer record so that it can charge you each month. It holds our internal IDs, the email address you enter at checkout and your saved payment method. You manage the plan, your card and your invoices on Stripe's page, which the app opens.
- What we keep: a purchase record with the product, the credits, the amount and currency, Stripe's payment, order, invoice or subscription references, its state and the times; for a monthly plan, its status and when the current period ends. We need these records to add your credits, handle refunds and keep our accounts.
Anonymous usage statistics
The Android app sends our server a summary of how it was used once a day, so that we can decide what to improve. The summary contains:
- counts of your actions on each day: lookups and where they came from (the reading app, as one of a fixed list of apps or "other"), dictionary hits and misses, translations by engine, words saved, edited and deleted, how quickly the popup opened, which features were used, and crashes;
- the size of your notebook, in ranges;
- a few settings: translation language and engine, definition language, search scope, sort order and theme;
- whether a dictionary is installed and whether the account and sync are signed in;
- whether this is the first summary from this installation, overall or this month, so that we can count installations without an identifier;
- your device's region setting, the app's language, the app version and how long the app has been installed (in ranges).
It contains no words or text from your notebook or your reading, no account and no identifier. Our server adds it to totals and does not keep the summary itself. The server looks up which country your IP address is in and then discards the address. IP geolocation by DB-IP, licensed under CC BY 4.0.
If we add anything to the summary, we add it to this list first. Turn it off in Settings → About → Share anonymous usage statistics.
App updates
Fusen for Android is installed directly, not from an app store, so the app checks for updates itself: when it starts, once a day and when you ask, it reads the latest version from our server (api.fusen.ink), sending only the app version, as every request does. If you choose to update, it downloads the new version from our download server (download.fusen.ink), which, like any website, sees your IP address. The app checks the file's checksum, package name, version and signature before Android asks you to install it.
Feedback by email
If you use Send feedback, your email app opens a draft that includes the app version, the Android version, the phone model and the app's language. We receive what you send, including your email address, and use it only to answer you.
This website
fusen.ink uses no cookies, no analytics and no tracking. Like any website, the service that hosts it (Cloudflare) processes your IP address and request details to deliver pages and protect the site from attacks.
Services you connect to yourself
These services receive information directly from your device. We do not receive it, and their own privacy policies apply.
- Google Drive sync. If you turn on sync, the app saves your notebook to the app-data folder of your own Google Drive, which only Fusen can see. It asks only for access to that folder (
drive.appdata), not to the rest of your Drive. Images saved with words go there too only if you turn on Sync images (off by default, set on each device); they use your own Drive storage. We do not have a copy. - Google sign-in. Google's sign-in tells the app your email address and gives our server a token that proves which Google account you are. We do not receive your Google password.
- DeepL, Anthropic or an OpenAI-compatible server, if you choose one as the translation engine with your own API key: the passage you translate is sent directly to that service under your agreement with it. A server you enter yourself may use plain http, which is not encrypted.
- Google Translate. If you tap the Google Translate button, the passage is handed to the Google Translate app.
- Dictionary downloads. The dictionary (JMdict) is downloaded from the Electronic Dictionary Research and Development Group's server (edrdg.org), or from a copy on our download server (download.fusen.ink) where the app offers one. Definition packs, such as the Chinese definitions, are downloaded from our download server. Each of these servers, like any website, sees your IP address.
Google user data
Fusen's use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. We use Google user data only to provide sync and sign-in. We do not use it for advertising, do not sell it, and do not use it to train AI models.
Service providers
We use these companies to run Fusen. They process information only on our behalf and as described here:
| Provider | What for | Where |
|---|---|---|
| Amazon Web Services (including Amazon SES) | Our server, database, backups, report images and sign-in emails | Tokyo, Japan |
| The AI providers listed above | The AI behind Fusen AI | As listed above |
| Stripe | Payments, when you buy credits or a monthly plan | United States and other countries |
| Cloudflare | This website, and app and dictionary downloads | Worldwide network |
The DB-IP database used to find a country from an IP address is a file on our server; nothing is sent to DB-IP.
Before a new service provider receives your personal information (for example another email service that delivers better in some countries, or another AI provider), we add it to this policy, and we tell you in the app before the change applies.
Where your information is processed
Our server, its database and its backups run on Amazon Web Services in Tokyo, Japan. Text and images sent to Fusen AI are processed by the AI providers in the countries listed above, and payments by Stripe, which operates in the United States and other countries. The United States has no single comprehensive national data-protection law; personal information is protected there by sector-specific federal laws and by state laws such as California's Consumer Privacy Act. If we add an AI provider in another country, we describe that country's data-protection system here first. Services you connect to yourself process your information where they operate.
How long we keep information
| Information | Kept |
|---|---|
| Account, credits, sessions | Until you delete your account. Sessions expire after 180 days unused. |
| Usage records | 400 days, or until you delete your account |
| Ratings of AI answers | 180 days, or until you delete your account |
| Reports | 180 days (images 181 days), or until you delete your account |
| Purchase records | At least 7 years after the year of the purchase, as tax law requires, also after you delete your account |
| Server logs | 30 days |
| Email sign-in codes | Until they expire (15 minutes) |
| Hashed IP addresses for email sign-in | 2 days |
| Anonymous usage statistics | Added to totals, which are kept |
| Text you translate or explain, images you ask Fusen AI to read | Not kept by us |
| Backups of our database | 35 days |
Backups. Our database is backed up continuously, and a copy is taken once a day. The backups are encrypted, are kept in the same Amazon Web Services region as the server (Tokyo, Japan), and are used only to recover from a failure or a mistake. Information we delete from the database, such as your account, credits, sessions, usage records, ratings and reports when you delete your account, can remain in these backups for up to 35 days before it is gone for good.
After you delete your account, we keep a small record for 35 days (a year if the account was blocked for abuse), so that deleting and recreating an account cannot be used to get extra free credits. It holds your Google account ID or a hash of your email address, the free credits left that month, the number of requests the AI declined, and whether the account was blocked. It contains no email address in readable form and no usage records. Deleting your account also deletes your customer record at Stripe, if you had one; purchase records stay as described above.
Your choices and rights
- Delete your account: in the app, Settings → Account and sync → Delete account, or without the app as described on our account deletion page.
- Turn off usage statistics: Settings → About → Share anonymous usage statistics.
- Delete synced data: Settings → Account and sync → Delete synced data deletes the notebook files and images sync keeps in your Google Drive.
- Remove Google access: in your Google account, Security → Third-party apps & services.
- Access, correction and deletion: you can ask us what we hold about you, and to correct or delete it, by emailing support@fusen.ink from your account's address. We answer within 30 days. Depending on where you live (for example under Japan's Act on the Protection of Personal Information or the EU's GDPR), you may have further rights, including to complain to your data-protection authority.
Security
All connections to our server use HTTPS. Session tokens are stored on our server only as hashes. Access to the server, the database and report images is limited to the developer, with multi-factor authentication.
Children
Fusen is not directed at children under 13, and we do not knowingly collect information from them. If you believe a child has created an account, contact us and we will delete it.
Changes
If we change this policy, we will update the effective date above. If a change affects what we collect, who receives it (for example a new AI provider) or where it is processed, we will also say so in the app before it applies.
Contact
The developer's name and address are provided without delay on request.