Tako Talk Privacy Policy
Effective October 9, 2026
Tako Talk is an AI-assisted Japanese-learning app operated by Tetsuya Maeda. This policy explains the information we process, why we use it, and the choices available to you.
Information we process
- Account information: a Firebase user identifier and information made available by Apple or Google sign-in, such as your name and email address.
- Learning information: preferences, saved vocabulary, review progress, companion progress, and memories or upcoming events that you approve. A minimized learning snapshot is synced across your devices.
- Conversation and audio content: messages, audio, and transcriptions are processed to provide AI conversation, speech, readings, and feedback. Raw audio and full conversation content are not placed in our server operational logs. Conversation history remains on your device unless you remove it.
- Journey speaking recordings: recordings made for immediate listen-and-compare practice remain in the app’s temporary storage, are never uploaded, and are discarded when you leave or background the lesson.
- Purchase information: Apple-signed transactions, product and entitlement status, and opaque account tokens used to verify and restore subscriptions and phrasebook purchases. In app versions with RevenueCat enabled, RevenueCat also processes an opaque support identifier, purchase, renewal and refund history, and technical request metadata for purchase support and revenue reporting. RevenueCat may also hold time-limited support access granted by our team. Signed-in users may have a RevenueCat record even without a purchase. To help our team identify your account and provide purchase support and time-limited access, our server shares the sign-in email address and display name available in Firebase Authentication with RevenueCat. This also applies to existing linked customer records. If you use Apple’s Hide My Email, this may be an Apple relay address. We also share an account classification, such as test, team, customer, or unclassified, to distinguish operational and test records in reporting. These contact details are used for account support, not advertising or marketing messages. We do not send RevenueCat your raw Firebase user ID, conversations, or audio.
- Security and operations: installation and app version information, App Attest and DeviceCheck proof, bounded usage counts, route, status, latency, and sanitized error categories used to protect and operate the service.
- Operational statistics: minimal measurement versions do not include Firebase Analytics or send app screen-view and interaction events. Our team uses aggregate counts from existing account and purchase records for service planning. Reports do not contain names, email addresses, account identifiers, conversations, or audio.
- Call troubleshooting: versions with this feature keep a bounded recent history of compact technical call summaries on your device. In versions where server collection is enabled, we automatically receive one compact technical summary per call, up to 2 KiB. We do not collect detailed event traces or retain separate long-term call-diagnostic aggregates. Reports include app version/build, OS version, broad device and audio-route categories, coarse call time, relative timings, counters, fixed result and issue categories, and random per-call diagnostic identifiers. They do not include recordings, conversation text, AI replies, memories, prompt content, names, email addresses, or names of connected audio devices. Our authenticated service and request records may allow these diagnostics to be associated with your account; they are not guaranteed to be anonymous. Access is limited to maintenance purposes. We do not use them for advertising or screen-interaction analytics.
- Crash diagnostics: Firebase Crashlytics processes crash traces, app version, device/OS details, and technical identifiers needed to diagnose reliability problems. We do not attach conversation, account, or profile values to crash reports.
- YouTube thumbnails and video playback: Journey loads video thumbnails from YouTube when you open a video example, sharing your IP address and request information. Thumbnail requests do not use stored cookies or a disk cache. The privacy-enhanced YouTube player loads only after you choose to play a reviewed example. During playback, Google/YouTube may receive your IP address, approximate location, device information, player interactions, advertising information, and diagnostics under Google’s policies. The player uses a non-persistent web data store, and Tako Talk does not download or cache YouTube audio or video.
How we use information
We use information to authenticate accounts, sync learning progress, personalize learning, generate AI responses and speech, verify purchases, prevent abuse, enforce service limits, fix crashes, and operate Tako Talk. Operational reports show account totals, accounts created in a reporting period that still exist, and purchase metrics. We do not treat sign-in counts as daily active users. These reports do not contain search queries or promotion-code text.
AI processing and your permission
Our third-party AI provider is Google, through the paid Gemini API. We collect the content you type or speak in AI features and send Google your microphone audio during live AI calls, messages, conversation history and transcriptions, text submitted for speech or reading generation, and words or conversation excerpts submitted for explanations and feedback. We also send your learning display name, learning preferences, and approved memories or plans when needed to personalize a response. Personal information you include in your conversations is part of this content.
Google processes this information to generate AI conversations, speech, transcriptions, readings, vocabulary explanations, reviews, and conversation stamps. Text and learning requests pass through our secure backend before reaching Google; live AI audio connects directly from the app to Google using short-lived credentials. Google may also receive network and technical request information, including your IP address for direct connections. Your sign-in email, purchase receipts, and account authentication credentials are not included in AI prompts.
Before sending this content, the app identifies Google (Gemini) and summarizes the shared data in an early onboarding step. You must select the initially unchecked consent checkbox, then choose Continue. The checkbox summarizes voice, chats and learning data shared with Google (Gemini). Continue is disabled until you select the checkbox. The AI & Privacy link opens the detailed disclosure, including audio, transcriptions, practice text, learning display name and preferences, and approved memories or plans; opening or closing it does not grant permission. You can leave setup without agreeing, and no AI content is sent. Existing users are asked once when updating to a version that introduces this permission. Signing in, completing earlier onboarding, and granting microphone access do not grant this permission. The choice is stored on your device and is not assumed from a synchronized learning profile. AI features remain off unless you agree. You can withdraw permission in Settings → Data & Privacy, which prevents new AI requests; withdrawal does not undo processing that has already occurred. Re-enabling AI shows the disclosure again. A new device or reinstall may require a new choice, and material changes to the recipient or shared data require renewed permission.
We require our AI service provider to provide the same or equal protection for shared personal data as described in this policy. For paid Gemini processing, Google’s Gemini API terms and Data Processing Addendum provide purpose limits, confidentiality and security commitments, and deletion obligations. Google states that paid-service prompts and responses are not used to improve its products or train its models. Google may log prompts and responses for a limited period for safety, abuse prevention, and required legal disclosures. This is separate from our server-log minimization and does not mean Google retains no content. Technical service data is also subject to the Google Privacy Policy.
Usage measurement
Minimal measurement versions disable Firebase Analytics regardless of a previous sharing preference, and do not show an analytics consent prompt or sharing switch. Earlier released versions may still collect optional Firebase Analytics events after you choose to share; their sharing setting remains in Settings → Data & Privacy. Updating the app does not delete previously collected data. Crash diagnostics, security operations, purchase processing, and learner-initiated YouTube playback described in this policy continue separately.
Call troubleshooting and optional reports
Technical call diagnostics help us investigate connection, microphone-processing, and playback problems. They are separate from Firebase Analytics and Crashlytics. A microphone permission does not itself authorize sharing a support report. You can review recent technical details in Settings → Call troubleshooting and choose to share a short summary. Sending feedback or sharing a report is optional. Opening the feedback form does not automatically attach your email address or diagnostic identifiers to its URL.
Local call summaries have a 7-day retention deadline and a maximum of 20 recent calls; earlier removal may occur when storage limits are reached. They are excluded from iCloud backup. The app removes expired reports on its next retention check; it cannot perform deletion while it is not running. Local reports are also cleared on sign-out or data deletion. Where server collection is enabled, call summaries have a 7-day retention deadline measured from the call, not a later retry. Cloud deletion processing can continue after that deadline, so it is not a guarantee of immediate physical erasure. Copies you deliberately export to another app or recipient are controlled by that recipient and its retention practices.
Firebase Crashlytics has a separate retention policy: Google describes beginning deletion of crash data after 90 days. The 7-day call-summary deadline does not apply to Crashlytics, independently required security records, or copies you choose to export.
Service providers and disclosure
Tako Talk uses Apple for sign-in, device integrity, and App Store purchases. It uses Google services including Firebase Authentication, Firestore, Analytics, Crashlytics, Cloud Run, Hosting, Gemini, and YouTube for account services, secure storage, diagnostics, hosting, AI processing, and learner-initiated video examples. App versions with purchase reporting enabled use RevenueCat for customer purchase records and subscription reporting. Providers process information under their own terms and privacy commitments. We may disclose information when required by law or to protect users and the service.
We do not sell personal information or track you across other companies’ apps and websites. The standard YouTube player may show ads under Google’s policies; Tako Talk does not operate its own advertising network.
Journey video playback is also subject to the YouTube Terms of Service and Google Privacy Policy.
Storage and international processing
For support access and additional conversation allowances, we record an opaque account identifier, the operator, the reason, the allowance and its expiry. Support allowance settings and their audit records are retained for up to 400 days, with deletion processed by our storage provider after the retention deadline. Daily operational snapshots contain aggregate counts without account identifiers and are retained for up to 400 days. Completed RevenueCat deletion-request evidence has a 30-day retention deadline and is removed through our conditional cleanup process; unfinished requests are kept until resolved.
Your account and synchronized learning snapshot remain until you delete them. Security challenges, usage ledgers, subscription status, diagnostic records, and fraud-prevention records are retained only as needed to provide and protect the service, meet legal obligations, and resolve disputes. Apple may retain purchase records independently. Our production backend is hosted in the United States, and Apple, Google, and RevenueCat may process information in other countries where they operate.
Your choices and rights
You can export or remove conversations and learning data in the app. You can delete your Tako Talk account from Settings after reauthentication; this removes the synchronized learning snapshot and Firebase account. In versions with RevenueCat enabled, the app submits a purchase-record deletion request before deleting your Firebase account. This request is processed separately; purchase and security records may be retained where needed for legal obligations or dispute resolution. Account deletion does not cancel an App Store subscription, which is managed separately through Apple.
Depending on where you live, you may have rights to access, correct, delete, or restrict certain processing. Contact us to make a request.
Security, changes, and contact
We use transport encryption, short-lived credentials, Apple device-integrity checks, access controls, and minimized server identifiers. No system is completely secure. We may update this policy as Tako Talk changes and will update the effective date or provide additional notice when required.
Questions or privacy requests: support@maepace.com