← Kotoba

Kotoba Privacy Policy

Effective September 29, 2026. Kotoba is operated by E KIMHENG, Japan. Contact: ekimheng.apps@gmail.com.

What stays on your iPhone

These are stored only on your device. We don’t keep copies. When you translate, the parts needed for that request are sent as described in the next section.

You can delete history and saved items in the app. Deleting Kotoba removes all of it. Like other app data, it may be included in your iPhone or iCloud backups under Apple’s terms.

Text you translate

Translation and Reply Coach use AI. Each request sends the following to our server:

For Reply Coach, this includes up to the last 12 turns of the practice conversation; for suggested replies, up to 3 earlier suggestions, so new ones don’t repeat.

Our server forwards this to DeepSeek, which generates the translation, vocabulary, and replies. We don’t save your text on our server, and our logs don’t contain it; they record only technical details such as response times and error types.

DeepSeek processes this data as our service provider. DeepSeek is based in China, and your text may be processed there. DeepSeek’s published terms don’t state how long it keeps text sent through its API, and they don’t promise not to use that text to improve its models. See DeepSeek’s Open Platform Terms of Service and DeepSeek’s privacy policy.

Please don’t enter information you want to keep private, such as passwords or health or financial details.

If you don’t want your text processed this way, don’t use translation or Reply Coach. The dictionary, handwriting, and flashcards work without sending any text.

Voice input

When you hold the record button, Kotoba uses Apple’s speech recognition to turn your voice into text. Apple may process the audio on its servers under Apple’s privacy policy. Kotoba doesn’t save recordings or send audio to us or to DeepSeek. The resulting text is handled like typed text.

Handwriting and the dictionary

The dictionary is built into the app and searches on your device. Handwriting recognition uses Google ML Kit, which recognizes your writing on your device; your drawings aren’t sent to us. ML Kit contacts Google to download recognition models and sends Google usage and diagnostic information to measure and improve its performance: device and app details, an identifier for the installation, the features and languages used, performance metrics, and errors. Google says this information isn’t linked to your identity or used for tracking. Kotoba includes no other analytics or advertising SDKs. See ML Kit terms and privacy.

Keeping the service fair and secure

Kotoba uses Apple’s App Attest to confirm requests come from the genuine app. Apple creates a key for your installation; our server stores its identifier, public key, and a signature counter for up to 120 days after last use. We use that identifier to count free translations each day, apply rate limits, and prevent abuse. Daily counts expire shortly after midnight UTC.

The app contacts our server when it opens, to prepare the service. These requests carry App Attest data and, if you subscribe, your subscription record, but none of your text.

Our hosting provider processes IP addresses to deliver requests and protect the service. If a request reaches our server without App Attest, we use its IP address for rate limiting that day.

Purchases

Apple processes Kotoba Pro purchases. We never receive your payment details or Apple ID. To unlock Pro features, the app sends Apple’s signed record of your subscription to our server, which confirms it with Apple. We use the subscription’s transaction identifier only to check that it’s active and to apply rate limits. Our server keeps it only in temporary memory, which is cleared whenever the server restarts, and in a one-minute rate-limit counter.

Service providers

ProviderRoleData
DeepSeekTranslation, vocabulary, and repliesText you send for processing
VercelHosts our serverRequests, IP addresses, technical logs
UpstashShort-lived storage for counters and App Attest recordsApp Attest identifiers and public keys, one-time security challenges, request IDs for daily counts, rate-limit counters
AppleSpeech recognition, App Attest, purchasesAs described above, under Apple’s policy
Google (ML Kit)On-device handwriting recognitionModel downloads, usage and diagnostic information

We require each provider to protect your data at least as well as this policy does and to use it only to provide its service to us, with two exceptions: Google uses ML Kit usage and diagnostic information to measure and improve ML Kit, as described above, and DeepSeek’s terms don’t rule out using your text to improve its models. These providers may process data in the United States, China, and other countries.

How long we keep data

Your translation textNot stored by us. DeepSeek’s terms don’t state a retention period.
App Attest recordsUp to 120 days after last use
Security challenges5 minutes
Daily usage counts and rate limitsUntil shortly after midnight UTC, or one minute for Pro rate limits
Server logsKept briefly for troubleshooting, under Vercel’s standard retention
Everything on your iPhoneUntil you delete it or delete the app

Your choices and rights

Depending on where you live, you may have rights to access, correct, or delete your personal data, or to object to its use. Because Kotoba has no accounts and doesn’t keep your text, we usually hold nothing that identifies you. To make a request, email ekimheng.apps@gmail.com, and we’ll reply within 30 days.

Children

Kotoba isn’t directed at children under 13, and we don’t knowingly collect their personal information.

Changes

If we change how Kotoba uses your data, we’ll update this page and its effective date. If the change affects what is sent for AI processing, we’ll also tell you in the app.