Privacy Policy

Version 2026-10-07 · Effective 2026-10-07

AI English Coach (the "App") is operated by Zhang Tao, an individual developer based in China ("we", "us"). This policy explains what personal information we collect, how we use, share, keep and protect it, and your rights.

At a glance

• Only after you explicitly agree to AI data sharing, practice audio and content are sent through our server to the named third-party AI recipients below. Microphone permission alone is not agreement. Recordings are not stored by default on our practice server; this does not describe provider retention.
• One of them, DeepSeek, processes and stores data in mainland China and may use it, de-identified, to improve its services; we cannot confirm that it protects your data to the same standard as this policy (see section 4).
• The retention job clears complete transcripts after 180 days; learning evidence and correction excerpts can remain until account deletion.
• The current capture path does not write raw recordings to storage. Any legacy recording-retention permission is separate from AI sharing; stored legacy audio, if present, is covered by the 30-day clean-up rule below.
• We do not sell your personal information, we do not track you across apps, and the App has no ads.
• You can export your data, withdraw consent or delete your account in the App.

1. What we collect

Sign in with Apple or Google. When you choose one of them, Apple or Google confirms who you are and sends us an account identifier and your email address (Apple may give a private relay address instead of your real one). We use them only to create your account or link the method to your existing account. We never receive your Apple or Google password, and we do not store your name or profile picture even if the provider shares them. If you sign in with Apple, we may also hold an Apple-issued token that lets us revoke the App's access to your Apple ID; we revoke it when you unlink Apple or delete your account. You can see and unlink your sign-in methods in Settings › Account (at least one must remain). Apple's and Google's own privacy policies apply to their sign-in services.

2. How we use it

• To provide the service: sign-in, daily missions, real-time conversation with the AI teacher, corrections, review, learning reports and your language profile.
• Subscriptions and limits: to check your subscription and apply the free daily voice allowance.
• Security and abuse prevention: rate limits, detecting misuse (for example, using the App as a general-purpose chatbot or trying to extract its instructions) and cost control.
• Improving the product: using allowlisted service statistics, not conversation text or microphone recordings. This revision has no recording-based product-improvement feature.
• Communicating with you: sign-in codes, service notices and replies to your feedback.
• Legal compliance: for example keeping transaction records and answering lawful requests.

We do not use your personal information for advertising, sell it, or combine it with other companies' data to track you. We do not use voiceprints or use your voice to identify you. Your "estimated level" is calculated automatically to plan lessons; it is not an official certification and has no legal effect on you.

3. Voice and independent AI data-sharing choice

1. Before personal practice content is forwarded, the App displays the current named recipients, data, purposes and their privacy-policy links. Choose Agree and continue or Not now. Nothing is preselected; accepting Terms or enabling the microphone does not grant AI sharing.
2. Audio travels over encrypted connections to our server and then the speech-recognition provider. Transcripts and necessary learning context go to the teacher / decision services. Teacher text, which can repeat your own words, goes to speech synthesis. Post-session analysis can send conversation context after you leave the practice screen, while a valid sharing grant remains in force.
3. The current server capture path does not persist raw microphone recordings. Transcripts and session snapshots are stored for corrections and reports, linked to your account. The retention job clears complete turn text / word timings and ended-session snapshots older than 180 days. Learning evidence, correction excerpts and reports can remain until account deletion.
4. Declining disables features dependent on sharing; existing lesson reading, progress and account controls remain available. Publisher-owned lesson narration and fixed synthetic connection tests do not contain your practice input.
5. Withdraw in Me → Privacy & consent → AI data sharing → Withdraw AI data sharing. The App stops capture and pending transmissions locally first; the server cancels affected queued analyses and checks consent at HTTP requests and outbound audio frames. Offline withdrawal remains blocked on this device and retries its consent notice when connected. Until the server confirms, other devices may still hold the old permission. Data already received by a provider is not automatically deleted by withdrawal; contact us for deletion requests.
6. A grant is account-specific and records a disclosure version. New recipients, destination changes or material changes to data/purpose require a new explicit choice. Legacy Terms, Privacy and voice-processing consents are not treated as AI agreement.
7. AI replies can be incorrect and should not be treated as professional advice. We do not use voiceprints to identify you.

4. Named AI recipients and other recipients

The following recipients were verified against the live server configuration on 6 October 2026, and each provider's terms were checked on 7 October 2026. The in-App disclosure is generated from callable providers, including eligible backups; undisclosed providers or unverified proxy destinations are blocked for personal practice data. New provider activation must be audited and re-disclosed first.

We do not separately attach your account email, installation identifier or internal account ID to these AI payloads. Personal information you say or type can still be present in the content. Providers also receive our server connection metadata and authentication credentials.

How each recipient keeps and uses data (per its published terms):

• Deepgram: every speech-recognition request we send opts out of Deepgram's Model Improvement Program (mip_opt_out=true). Deepgram's "Your Data" documentation and Model Improvement Program page say that opted-out requests do not have their audio or transcripts stored after the result is returned, and are not used to train models; Deepgram keeps request metadata and usage logs that contain no audio or text. This comes from Deepgram's documentation; its terms of service do not themselves contain this commitment, and we do not have a data processing agreement with Deepgram.
• DeepSeek: DeepSeek processes and stores data in mainland China. Under its Open Platform Terms of Service and the Terms of Use they sit under, DeepSeek may use inputs and outputs, after encryption and de-identification, to improve its services and technology. DeepSeek does not state how long API data is kept, publishes no data processing agreement, and its terms are governed by the law of the People's Republic of China. We cannot confirm that DeepSeek protects your data to the same standard as this policy, so we do not make that claim. What you should know: your practice transcripts and learning context are sent to mainland China and may be kept there; DeepSeek may use them, de-identified, to improve its services; withdrawing consent does not automatically delete data DeepSeek has already received. Please don't say or type sensitive information such as ID numbers, health or financial details during practice. If you don't want your content sent to DeepSeek, choose "Not now" or withdraw at any time; AI conversation, corrections and learning reports will then be unavailable, while lesson reading, progress and account features keep working.
• ElevenLabs: we use a paid ElevenLabs plan and have switched off, in the account's "Data use" setting, permission for ElevenLabs to train on our content. Under its Terms of Service this applies to content sent after the change. Its Data Processing Addendum applies to our account but sets no retention period for this kind of plan, and zero-retention mode is only offered to enterprise customers. ElevenLabs may therefore keep the teacher text it receives (which may repeat your words) and the audio it generates; we do not promise zero retention or a deletion period for it.
• TypeSafe AI: under its Master Customer Agreement and Data Processing Addendum, TypeSafe does not use customer data to train models without our prior consent; the agreement has confidentiality terms but sets no retention period. The agreement lets TypeSafe use technical telemetry about service use (such as logs and statistics) to improve its products.

For Deepgram (with the opt-out), ElevenLabs (paid plan, training off) and TypeSafe, these terms and settings support saying that they do not use the data to train models (for Deepgram, per its documentation). That does not mean they process it only to provide their service to us: for example, ElevenLabs' terms let it use content it receives for purposes such as improving its services, and TypeSafe may use technical telemetry to improve its products. None of them commits to a specific deletion period. DeepSeek's terms do not support that statement, which is why it is described separately above.

Apple / Google authenticate you if chosen. Apple handles in-app purchases. Hosting and sign-in email infrastructure also process data needed for those functions; these are separate from the AI-sharing permission. Their actual product configuration and retention must be documented before publication. We may disclose information where legally required, to protect users, or in a business transfer subject to applicable safeguards.

5. How long we keep it

6. Your rights

• Access and export: Me → Legal → Export my data (a JSON file).
• Correction: change learning settings in the App, or contact us to correct anything else.
• Delete your account: Me → Account → Delete account. Deletion is scheduled after a 30-day grace period; contact us if it has not completed. Cancel any subscription separately in your Apple ID settings.
• Withdraw AI sharing: use Me → Privacy & consent → AI data sharing. You need not delete your account. It blocks future sharing; it does not automatically erase previously received provider data. Legacy recording-retention permission can be turned off separately in Settings.
• Other rights: depending on where you live, you may also restrict or object to processing, port your data, get an explanation, or complain to a regulator (section 9).

We reply within 15 working days, or sooner where the law requires. We may need to verify your identity first to protect your account.

7. Children

The App is only for people aged 18 or over, and you confirm your age at sign-up. We do not knowingly collect minors' personal information; if we learn a minor is using the App, we disable and delete the account. Contact us if you believe we hold a minor's information.

8. Where data is stored, international transfers and security

• Location: our servers are currently in the Philippines (Manila). The live host reports Alibaba Cloud region ap-southeast-6 (Manila). The API routing code sg is a logical service region, not proof of a Singapore data center. The AI recipients are in the United States (Deepgram, ElevenLabs, TypeSafe) and mainland China (DeepSeek); see section 4.
• International transfers: AI providers may be located outside your country or region (for example in the United States). Cross-border safeguards and applicable API processing terms must be verified before this revision is published. We do not claim an executed Standard Contractual Clauses agreement without evidence.
• Security: TLS in transit; application-layer encryption of emails; short-lived, rotating sign-in credentials; AI provider keys only on our servers; application-log redaction for known conversation/audio fields and email/phone masking; role-based admin access with audit logs. If a personal data breach occurs, we will notify you and regulators as the law requires.
• Cookies: the App uses no cookies, advertising identifiers or third-party ad or analytics SDKs, and this page sets no cookies. Sign-in credentials stay in your device's Keychain.

9. Regional information

9.1 Mainland China (Personal Information Protection Law, PIPL)

• This policy is our personal information handling rules; contact us at the email in section 11.
• Separate consent: the AI-sharing screen provides the named recipients, data and purposes with Agree and continue / Not now. It is separate from microphone permission and Terms acceptance. Voice may be sensitive personal information; additional regional consent and transfer requirements require legal verification before offering the service in that jurisdiction.
• Cross-border processing: the current server is in Manila, Philippines; AI data goes to the named recipients above: DeepSeek in mainland China, and Deepgram, ElevenLabs and TypeSafe, which are US companies. The transfer mechanism and recipient contractual contacts remain publication checks. Do not infer a legal cross-border basis from the service's logical region label.
• You have the rights to access, copy, correct, supplement and delete your information, withdraw consent, close your account, obtain an explanation and transfer your data.
• AI-generated content is clearly labelled in the App.

9.2 EEA, UK and Switzerland (GDPR)

• Controller: Zhang Tao, an individual developer based in China; contact details in section 11.
• Legal bases: contract (account, basic learning service, subscriptions); explicit choice for AI sharing, with consent and any additional legal basis/requirements assessed for each applicable market; legitimate interests (security, abuse prevention, service statistics that use minimal data); legal obligation (transaction and tax records).
• You have the rights of access, rectification, erasure, restriction, portability, objection and withdrawal of consent, and you may complain to the data protection authority in your country.
• The applicable processor agreements and lawful safeguards for transfers outside the EEA require verification before distribution in those markets.

9.3 Thailand (PDPA)

We process your personal data on the bases of contract, consent and legitimate interests. You have the rights to access, obtain a copy, correct, erase, restrict, object, port your data and withdraw consent, and to complain to the Personal Data Protection Committee (PDPC). The applicable cross-border mechanism must be verified before distribution in Thailand.

9.4 California (CCPA/CPRA)

In the last 12 months we collected: identifiers (email, account and install identifiers), commercial information (purchases), internet activity (usage data), audio information (voice) and inferences (learning profile), for the purposes in section 2. We do not sell your personal information and do not "share" it for cross-context behavioral advertising. You have the rights to know, delete and correct your personal information, and we will not discriminate against you for using them. Use the in-App options or email us; you may use an authorized agent.

10. Changes to this policy

We may update this policy. The version and effective date are shown at the top. For material changes we will tell you in the App and ask for your agreement again before you continue.

11. Contact us

• Operator: Zhang Tao, an individual developer based in China
• Email: vsv.zhang@gmail.com

See also the Terms of Service.