Last updated: October 25, 2026
Effective date: October 25, 2026
This is the English translation of the privacy policy for the version of MiaoYan AI Keyboard downloaded from our official website (miaoyanzhixin.com / miaoyan.co, including the Diệu Ngôn edition). The Chinese version at /privacy.html prevails in case of any inconsistency. The version distributed on Google Play has its own policy at miaoyan.co/privacy-en.html.
Shenzhen Chainfun Technology Co., Ltd. ("we", "us") understands how important your personal information is to you. As required by law, we take appropriate security measures to keep your personal information safe and under control.
| Type of information | Purpose | Where it is stored | Retention |
|---|---|---|---|
| Phone number | Account registration and sign-in | Cloud servers | For as long as your account exists |
| Chat content you enter | Used as context for the AI to generate replies | Sent encrypted to our servers and forwarded to the AI model you selected for processing | Kept on the server only while the session is being processed; you can clear conversation history in the app |
| AI-generated reply text | Providing the AI writing service | Cloud (usage counts only, not the text itself) | 30 days |
| Contact / chat partner information | Entered by you, used to personalize AI replies | Cloud servers | For as long as your account exists |
| Guest identifier and the IP address it was issued from | Tracking the free quota before you sign in; limiting how many identifiers are issued per IP per day to prevent abuse of the free quota | A random identifier issued by our server, stored on cloud servers together with the IP address at the time it was issued; we do not read Android ID, OAID or other native device identifiers | Automatically deleted after 90 days |
| Clipboard content | Quickly pasting the other person's message into the keyboard | Read temporarily on the device only, not uploaded | Not stored |
| Screen-scan screenshots | Recognizing the conversation on screen to suggest replies | Sent to our servers and forwarded to an image-recognition model | The original image is not kept; only a record of the recognition call is kept |
| Documents and materials you upload | Building personal / business profiles for the AI to use when needed | Cloud servers | For as long as your account exists; deleted together with your account |
| Voice recordings (microphone audio) | Hold the space bar to talk: speech-to-text, after which AI adds punctuation and fixes grammar in the recognized text | Streamed encrypted to our servers while you speak and forwarded in real time to a speech-recognition service; recordings are not kept. In the Vietnamese edition (Diệu Ngôn), voice goes through our overseas server to OpenAI (United States) and does not pass through mainland China | Not stored |
| Chat partner name at the top of a chat screen (optional feature) | Automatically recognizing who you are chatting with and switching to that chat partner | Only after you turn on "Accessibility – MiaoYan" in system settings yourself, the app reads the name shown at the top of chat screens in WeChat, QQ, Zalo, WhatsApp, Messenger and Telegram; matched on the device only; chat content is not read and nothing is uploaded | The name itself is not stored; only the "chat screen ↔ chat partner" links you have confirmed are saved on the device |
| Live text preview during a screen scan | Showing recognized text in real time while scanning | Recognized offline on the device, not uploaded | Cleared when the scan ends |
| Chat partner profile (organized by AI) | Organizing scanned chats into a profile the AI can refer to when replying | When organizing, the newly added chat content is sent encrypted to our servers and forwarded to the AI model you selected; the result is stored on the device only | Deleted when you uninstall the app or clear its data |
About "Automatically recognize the chat partner" (Accessibility service): this feature is off by default, is only offered in the website versions, and must be turned on by you under system "Settings → Accessibility". Once on, the app only reads the name at the top of chat screens in the chat apps listed above, to match it on the device against chat partners you created. It does not read, store or upload any chat content, and it never taps or types anything on your behalf. You can turn it off at any time in the same place, and it stops immediately.
About "Reply directions": when the keyboard receives the other person's message, it automatically offers a few reply directions (such as agree, politely decline, change the subject) for you to choose from. To generate them, the other person's message, excerpts of the profile you organized for that chat partner, and the recent conversation are sent to our servers in mainland China and handled in the same way as the existing "AI reply"; you can turn this off under "Settings → Reply directions". Location (optional): only after you tap "Use my location" and agree in the system prompt will the keyboard read your current location once, when the other person asks where you are or about a meal; it does not track location in the background. Coordinates and address text stay on your phone and are filled into the reply by your phone; they are not uploaded to our servers. Our servers only know whether a location is available. You can revoke the location permission in system settings at any time.
Our AI reply and image-recognition features are provided by calling compliant third-party large-model services deployed in mainland China (DeepSeek, Alibaba Cloud Qwen, Moonshot AI Kimi); speech-to-text is provided by Alibaba Cloud Model Studio real-time speech recognition (with iFLYTEK speech recognition as a fallback when it is unavailable; both are called from our server). The chat context and screen-scan screenshots you submit are first de-identified on our server and then sent to the model services above to generate replies; they are not sent to model services outside mainland China. Voice recordings are used only for transcription; audio is not stored. We do not use your input data to train models.
All AI-generated replies are suggestions only; you alone decide what to send.
| Service | Purpose | Data transferred | Location |
|---|---|---|---|
| DeepSeek | AI text generation | De-identified conversation context | Mainland China |
| Qwen (Alibaba Cloud) | AI text generation, screen-scan image recognition | De-identified conversation context, screenshots | Mainland China |
| Kimi (Moonshot AI) | AI text generation | De-identified conversation context | Mainland China |
| Alibaba Cloud SMS | Sending SMS verification codes | Phone number | Mainland China |
| Alibaba Cloud Model Studio real-time speech recognition | Real-time speech-to-text | PCM audio stream (transferred only, not stored) | Mainland China |
| iFLYTEK speech recognition (fallback) | Speech-to-text when real-time recognition is unavailable | PCM audio (transferred only, not stored) | Mainland China |
| OpenAI real-time transcription (Vietnamese edition only) | Speech-to-text in the Vietnamese edition | PCM audio stream (transferred only, not stored), relayed through our overseas server | United States |
| Tavily Search | Finding ice-breaker topics, analyzing link content | Only search keywords or the web address you provide, no conversation content | Outside mainland China |
We protect your data with SSL/TLS encrypted transmission, database access control and other technical measures. All cloud data is stored on compliant servers in mainland China.
This service is intended mainly for adults. If you are under 14, please use it under the guidance of a parent or guardian.
For privacy questions, please contact us:
Email: jetty@ichainfun.com (main company email)
Alternative: luojianming@miaoyan.co
We may update this privacy policy from time to time. The updated version will be published on this page.