Privacy Policy
Version: 1.3. Updated: 2026-11-16. Effective upon formal publication.
MySmartLED is provided by 广州奇景照明科技有限公司. Privacy and support: mysmartled@qq.com.
Scope and choices
This policy covers MySmartLED and its cloud services. Basic lighting control uses Bluetooth on your phone. Account sync, AI assistance and support requests involve separate cloud processing. You choose whether to use these features. Declining an optional permission does not prevent features that do not need it.
Bluetooth and local control
To discover, connect to and control lights, the app processes advertising names, Bluetooth identifiers, connection status, capabilities and control settings. An identifier may be a MAC address on Android or a system-assigned identifier on other platforms. Names, rooms, groups and scenes can be stored locally; relevant information may be uploaded when you enable sync or use AI or support.
Accounts and cloud sync
When you choose third-party sign-in, we receive verification credentials, an account identifier and the name, avatar or email returned with your authorization. We do not obtain your third-party password. Enabling sync uploads and stores light, room, group and scene settings; disabling it only stops future sync. Using AI without signing in also creates a persistent server-side guest identifier linked to conversations; this is not fully anonymous.
AI text and voice
Sending an AI text request passes your input, relevant conversation context and light identifiers, names, rooms, states and capabilities through our server to DeepSeek. The server stores conversations, responses and necessary invocation logs. Android main AI voice and transcription requests upload recordings to Alibaba Cloud Model Studio. iOS first uses on-device system recognition and uploads audio only when falling back because local recognition is unavailable or fails. Cancelling a recording does not start recognition; choosing transcription may already trigger recognition, while intent analysis starts only when the text is sent. Keeping original voice stores recordings in phone caches for playback; switching it off does not erase earlier caches. The Android offline recognition trial works locally after downloading models from Hugging Face; this does not make the main AI offline.
Local audio and permissions
Music effects read selected local audio; microphone effects analyse sound on the phone. Those local analyses do not themselves upload audio; AI recognition is separate. Bluetooth/nearby-device access supports discovery and connection. Android location declarations support Bluetooth compatibility, not geographical tracking; background location is not requested. Microphone access supports recording or effects you start. Android ongoing effects can use a foreground service and notification; stopping them stops that capture. Pickers or relevant system permissions read selected media/files. Network access supports requested cloud features, firmware and model downloads. iOS speech permission supports system recognition. Revoke permissions in system settings; refusal affects only features that need them.
Support and operational records
Support submissions include your description, selected images/videos, contact details, selected light name/Bluetooth identifier/firmware, phone model, OS and app version displayed on the submission page. Attachments upload before ticket creation; a failed ticket may leave uploaded files on the server, for which you can request deletion. Firmware checks and result reports process product identifiers, current/target versions, platform and update status to match updates and diagnose failures. Network services process request IP, time, errors and account or installation identifiers for operation and abuse prevention. Do not submit passwords, unrelated private material or others’ information without authority.
Recipients, locations and safeguards
Our cloud service is currently hosted in Hong Kong, China. DeepSeek and Alibaba Cloud Model Studio handle relevant requests through their mainland China services. Sign-in connects to the chosen provider; Android trial model downloads connect to Hugging Face, whose infrastructure includes the United States and other regions. Appendix II identifies data and transfer methods. Where required, we must give cross-border notices, obtain separate consent and implement applicable safeguards; reading this policy is not transfer consent. We use account authentication, administrator roles and access scopes and continue improving transport and file protection. No system is completely secure; security incidents require mitigation and legally required notification to users and authorities.
Retention and deletion
Account and sync settings are retained while the account exists and the service purpose remains necessary. Account deletion clears linked accounts, sync snapshots, AI sessions/call records and support database records. Server attachment files are not physically erased just by database deletion; request their removal through our email. AI sessions are retained as needed for history and related issues and support server-side deletion. Call logs are retained only as necessary for diagnosis, abuse prevention and legal obligations. Cloud speech audio is used to process recognition requests; our backend does not save it to disk as audio history. Phone voice caches and provider processing are separate. Tickets and attachments are retained for support, necessary follow-up and disputes; after the purpose ends or a valid deletion request is verified, they should be deleted or anonymised. Legally retained or temporarily undeletable backups/security records must have processing restricted to that purpose. Provider API terms determine provider retention; consumer policies are not a zero-retention guarantee.
Your rights
Signed-in users can initiate deletion under My → Account details → Delete account. Uninstalling, signing out or disabling sync does not delete cloud data. Guests can delete AI sessions and email us to remove guest identity and linked records; contact us before uninstalling or losing credentials. We verify only information needed to locate records, without requesting passwords or complete access tokens. You may request access, copies, correction, deletion or consent withdrawal, and, where applicable, restriction, portability or objection, and complain to a competent regulator. Withdrawal does not affect earlier lawful processing; you may stop AI requests and continue manual control. Within 15 working days of receipt we will verify and handle requests or explain lawful reasons for refusal; any shorter applicable legal deadline prevails.
Children and changes
The product is not directed to children and does not actively seek children’s data. Users below their local age of independent consent require lawful guardian consent and guidance; mainland China gives special protection to information of children under 14. If required guardian consent is absent, the relevant processing should stop and data be deleted as required by law. Guardians may contact our email. Material changes to purposes, data or recipients will receive prominent notice and renewed consent where required; continued browsing or preselected boxes are not new authorization.
Appendix I: Personal Data Collection List
Lights and local settings
Data: advertising names, Bluetooth identifiers, status, capabilities, colors, brightness, rooms, groups and scenes. Sources: Bluetooth discovery, light responses and your actions. Purpose: discovering, connecting, controlling and managing lights. Timing: when the relevant feature is used; basic control is primarily local.
Accounts and sync
Data: provider account identifiers, authorization credentials, returned profile details and settings you enable for sync. Purpose: sign-in verification, account management and restoring settings across devices. Timing: sign-in or enabled sync.
AI conversations
Sending an AI text request passes your input, relevant conversation context and light identifiers, names, rooms, states and capabilities through our server to DeepSeek. The server stores conversations, responses and necessary invocation logs. Android main AI voice and transcription requests upload recordings to Alibaba Cloud Model Studio. iOS first uses on-device system recognition and uploads audio only when falling back because local recognition is unavailable or fails. Cancelling a recording does not start recognition; choosing transcription may already trigger recognition, while intent analysis starts only when the text is sent. Keeping original voice stores recordings in phone caches for playback; switching it off does not erase earlier caches. The Android offline recognition trial works locally after downloading models from Hugging Face; this does not make the main AI offline.
Local audio effects
Data: selected audio and microphone sound signals. Purpose: rhythm analysis and lighting commands. Timing: when music or phone-microphone effects are enabled. Local analysis does not itself upload the original audio.
Support requests
Support submissions include your description, selected images/videos, contact details, selected light name/Bluetooth identifier/firmware, phone model, OS and app version displayed on the submission page. Attachments upload before ticket creation; a failed ticket may leave uploaded files on the server, for which you can request deletion. Firmware checks and result reports process product identifiers, current/target versions, platform and update status to match updates and diagnose failures. Network services process request IP, time, errors and account or installation identifiers for operation and abuse prevention. Do not submit passwords, unrelated private material or others’ information without authority.
Operation and rights
Data: request IP addresses, times, errors, account or installation identifiers and necessary security records. Purpose: network services, authentication and abuse prevention. Retention, recipients and deletion are explained in the privacy policy and third-party list. Data not actually processed is not represented as collected.
Firmware and model downloads
Data: product identifier, current/target firmware, platform, update status, download IP and time. Purpose: matching/distributing updates and diagnosing results. Trigger: checking, downloading and reporting updates or actively downloading offline speech models; see Appendix II for the model source.
Appendix II: Third-Party Data Sharing List
Scope of this list
This list covers used backend processors, sign-in components and model download sources. Entity details below come from public provider terms; contracting entities and API terms must match the actual service account. Open-source UI, networking and database code does not itself send personal data to its authors. We do not sell personal data or use app-collected data for personalized advertising.
DeepSeek · AI intent recognition
Recipient: Hangzhou DeepSeek Artificial Intelligence Co., Ltd.; privacy contact: privacy@deepseek.com. Method: our backend API. Data: text/transcripts, necessary context and relevant light names/identifiers/rooms/states/capabilities. Purpose and trigger: interpreting your submitted AI requests and producing replies/control proposals. Original speech is handled by the speech provider, not sent to this intent endpoint as part of the text request.
Alibaba Cloud Model Studio · Speech recognition
Recipient: Alibaba Cloud Model Studio. Its public agreement effective 2026-09-02 names 通义云启(杭州)信息技术有限公司 as service entity, using cloud resources from 阿里云计算有限公司. Method: our backend API. Data: audio requiring cloud recognition, format and necessary recognition parameters. Purpose: speech-to-text. Trigger: Android main AI voice/transcription and iOS local-recognition fallback. Contact channels and data conditions are in the official agreement/privacy notice. Microphone permission alone does not upload audio.
Google sign-in components
Recipient: Google sign-in (Google LLC; regional entities in its official policy). Method: Credential Manager/Google Identity or equivalent platform components. Data: credentials, account identifier and authorized profile information; the login service also handles network, device and service records under its policy. Purpose and trigger: identity verification when you choose Google sign-in. Its official privacy policy provides contact and rights channels.
WeChat OpenSDK / Sign in with Apple
Recipients: WeChat Open Platform (Tencent) and Apple sign-in (Apple Inc.; regional entities in its policy). Method: WeChat OpenSDK or system login API. Data: authorization credentials, identifiers, authorized profile and device/network details needed by components. Trigger and purpose: signing in through your selected, authorized option. iOS WeChat callbacks may use SDK clipboard interaction for login redirects, not unrelated clipboard collection. Sign-in is distinct from local speech recognition; on-device iOS recognition itself does not upload audio to our cloud. See each official policy for contact and rules.
Tencent Cloud · Hosting and file storage
Recipient: Tencent Cloud; its public statement identifies 腾讯云计算(北京)有限责任公司. Method: server hosting. Data: account/sync information, AI requests/logs, support/attachments and operational information submitted to the platform. Purpose and trigger: processing corresponding network requests, hosting and storage. Current support attachments are stored as files on the hosted server. Contact channels appear in its official statement.
Hugging Face · Offline model downloads
Recipient: Hugging Face, Inc. Method: direct phone requests for model and tokenizer files. Data: download IP, path, time and necessary HTTP request information. Purpose and trigger: serving files when you download/retry models in the Android offline trial. This download does not upload recordings or AI chats to Hugging Face; trial recognition runs locally after download. See its official policy for contact channels.
Other providers and updates
This app does not separately train models on your submissions. Actual API data agreements govern DeepSeek and speech providers’ retention, use and subprocessors. Public privacy links explain providers and do not replace contract verification or guarantee that providers never retain data or train on it. Before enabling backup models or new recipients, update this list and complete required notice and consent.
- DeepSeek
- Alibaba Cloud Model Studio
- Alibaba Cloud Model Studio · API
- WeChat OpenSDK
- Apple
- Tencent Cloud
- Hugging Face
AI Feature Notice
Text requests
After you send, your text, relevant conversation and light names, identifiers, rooms, status and capabilities go to our service and then to the relevant AI service for intent recognition. Results may propose lighting actions; actions requiring confirmation need your approval.
Voice requests
Recording uses the microphone. Android main AI voice/transcription uses Alibaba Cloud Model Studio; iOS prefers on-device recognition with cloud fallback when unavailable. Cancellation stops recognition; transcription can require uploading audio, while sending the resulting text passes it to DeepSeek. The keep-original setting controls future phone caches only. System microphone permission is not third-party sharing permission; before first sharing, disclose recipient, purpose and data and obtain required consent.
Local control and alternatives
Final lighting commands use local Bluetooth. You may control lights manually or use text instead of recording. Phone-microphone effects differ from AI voice requests and analyze sound locally.
Retention, recipients and rights
The server stores conversations and necessary call logs; this policy describes recipients, locations, caches and deletion. Before first sharing with DeepSeek or speech services, separately explain that processing and offer an explicit accept/refuse choice, obtaining separate consent where required. Refusal still allows manual control. Stop requests, delete sessions or email us to withdraw authorization; the operator must implement requests and provider restrictions.
Feedback Upload Notice
Information submitted
When you start submission or attachment retry, selected files upload first, followed by ticket content, contacts and the light/phone information displayed on the page. Unselected attachments do not upload. A failed ticket does not mean uploaded attachments have been removed; contact us to clear them.
Purpose and recipients
The operator and necessary hosting or storage processors handle data for troubleshooting, support replies and ticket management under the privacy policy. Do not include passwords, unrelated private images or information about others without permission.
Your choices
Optional details or attachments must not become mandatory for general feedback. Reading this notice does not check an agreement box; going back or cancelling does not submit a ticket. Missing diagnostics may require follow-up questions.
Retention and contact
Support records and attachments are retained for ticket handling, follow-up and necessary disputes. Email us for access, correction or deletion. Server files, database records and backups require separate cleanup; deletion of account records alone does not prove all copies were removed.
Processing grounds and sensitive information
We rely on grounds permitted by applicable law: necessary processing to provide requested account, sync and support services; affirmative choices for optional cloud AI, permissions and sharing requiring consent; and applicable legitimate interests or legal obligations for necessary security/legal records, limited to their purpose. Audio, photos and free-form input can contain sensitive information; avoid unrelated material. Sensitive data, transfers and other activities requiring separate consent must receive additional contextual notice and consent rather than blanket authorization through this policy.