At myFirst, we believe technology should support childhood, not consume it. We design products that give children room to explore, create, and express themselves while keeping clear boundaries around what technology can do, what it collects, and how it responds when something is not appropriate.
Because magiBox is designed for young children, we take the safety and privacy of its AI-powered generative features seriously. Safeguards are built into the experience to help prevent inappropriate requests from generating or printing content, while our approach to data is designed to limit what is collected, stored and retained.
The following Q&A reproduces information provided during the magiBox child-safety and data-protection review.
1. Will voice recordings be saved?
Response: No, they will not be stored. They are deleted immediately after use.
2. Will AI-generated images and stories be stored in the cloud?
Response: No, they will not be stored in the cloud. They are only stored locally in the user’s device.
3. Will prompts entered by users be used to train external AI models?
Response: No, they will not be used for model training.
4. How does the system handle inappropriate prompts entered by children?
Response: A prohibition pop-up reminder will be shown, and the corresponding function will be blocked from being triggered.
5. What is the exact data retention period for voice recordings, prompts, AI-generated images and stories? Can they be stored indefinitely?
Response: Only AI-generated images are saved. Other data is not stored on the server and is deleted immediately after use. Printed image results will be displayed in the app, retained for six months, and can be manually deleted. The image is only stored locally, and are not stored in the cloud.
6. Can parents delete a single record, a child’s personal profile or the entire account? After a deletion request is submitted, how long will the cloud take to completely remove the corresponding data?
Response: The cloud does not record data, and use of the magiBox and the AI features does not require a child account, so no personal information is recorded. Printed results displayed in the app can be deleted by long-pressing the corresponding item.
7. What personal data does the system collect? Please list all categories, including device identifiers, biometric information and usage behaviour. Does the collection follow the principle of data minimisation?
Response: No personal data is recorded. magiBox does not require children to create personal profiles or provide identifying personal information to use these features.
8. Will children-related data be sold, rented or shared with third parties, such as advertising or data analytics providers? If so, what conditions must be met and what authorisation is required?
Response: Since no personal data is obtained, such situations will not occur.
9. Which third-party AI and cloud service providers are integrated?
Response: magiBox uses Gemini, OpenAI and Tongyi Qianwen (Qwen).
10. How does magiBox handle harmful content, such as violence, self-harm, sexual content or hate speech?
Response: Requests that trigger applicable safety restrictions are blocked before generation. When a request is refused, magiBox displays “Oops! Try another idea,” and no image or story is generated or printed.
11. If crisis-risk content such as self-harm is identified, what is the complete handling process, and will guardians be notified?
Response: Requests identified as prohibited or harmful are blocked, and no corresponding content is generated or printed. magiBox does not provide guardian notification for these interactions because the service does not maintain an identified child profile associated with the request.
12. For children under 13, what is the parental authorisation process? Is verifiable parental consent required before collecting children’s personal information?
Response: Children’s information is not collected.
13. Is there an opt-out option to prevent children’s data from being used for AI model training? Is this option separated from service authorisation and not bundled with it?
Response: User prompts and generated content are not used to train external AI models.
14. Will targeted advertisements be pushed to child users? Can parents completely turn off advertisements and behavioural tracking?
Response: No. magiBox does not serve targeted advertising to children.
15. How are voice prompts protected while they are being transmitted?
Response: Voice prompts are protected with TLS encryption over WebSocket and AES-256-CCM encryption over Bluetooth during transmission.