Child Safety

1. A toy-like experience, not an open assistant

The robot is instructed to answer briefly, warmly, and in age-appropriate language. It does not browse the open internet, contact other people, expose a social feed, or ask a child to build a personal profile.

Silly jokes, imaginative play, and friendly teasing can be part of the fun. The safety design is intended to avoid making a child feel interrogated or punished while maintaining clear boundaries around adult or harmful topics.

2. Topics we restrict

Our safeguards are designed to block, avoid, or gently redirect discussion involving:

• Sexual content, nudity, pornography, grooming, romantic or adult relationship themes, and sexual-orientation or gender debates that are not appropriate for a young-child toy.
• Alcohol, illegal drugs, intoxication, smoking, vaping, and instructions for obtaining or using restricted substances.
• Graphic violence, killing, self-harm, suicide, weapons, dangerous challenges, and instructions that could cause injury.
• Hate, racism, dehumanizing stereotypes, bullying, severe profanity, threats, or targeted harassment.
• Requests for personal contact details, precise location, passwords, payment details, private images, or secrets that should be kept from a parent.
• Medical, legal, financial, or emergency guidance that should come from a qualified adult or professional.

3. How the safeguards work

Safety is applied in layers: the robot’s core instructions, input screening, output screening, limited response length, restricted conversation memory, and local fallback responses. If a topic is not suitable, the robot should change direction or suggest asking a parent or another trusted grown-up.

We tune the experience so harmless childlike silliness—such as jokes about smelly socks—can remain playful. Context and speech recognition are imperfect, so a harmless word may occasionally be misunderstood or a safe request may be redirected.

4. Voice and privacy protections

Kenzi does not store microphone recordings or persistent conversation transcripts. Audio is processed to understand the request and create the reply. Up to two recent turns may stay temporarily in memory for up to 15 minutes to maintain context and are not written to our conversation database.

Children should still be reminded not to share their full name, address, school, phone number, passwords, photographs, or family secrets with any connected toy or online service.

5. The parent or guardian’s role

Use device-level parental controls, keep purchase approval enabled, and choose supervision appropriate to the child’s age and needs. Listen to some early conversations together so the child understands that the robot is an AI character and can make mistakes.

If a response feels inappropriate, stop the interaction, reassure the child, note the approximate time and device platform, and report it to us. Do not send us a child’s unnecessary personal information when reporting a concern.

6. Urgent situations

Talking Robot cannot assess or respond to an emergency. If a child mentions immediate danger, abuse, self-harm, or another urgent concern, a trusted adult should act directly and contact appropriate local emergency or child-protection services. Do not wait for or rely on an AI response.

7. Continuous improvement

We review safety prompts, blocked-topic handling, speech misunderstandings, and reported incidents as the product evolves. Changes are tested with the goal of preserving a warm, funny robot personality while strengthening protection against unsuitable content.