Child Safety Policy

PHASE B — DOCUMENT 5: CHILD SAFETY POLICY

Status: draft for review. This is the public statement of how KidChat responds when a conversation raises a safety concern. It must describe only what exists.

A. WHAT WE PROMISE AND WHAT WE DO NOT

We promise to design for safety, to respond to risk signals with age-appropriate guidance and verified real-world resources, and to have a named person review the most serious events. (C-13, C-14)

We do not promise, and will not say, that we monitor conversations in real time, that we always detect harm, that we always block harmful content, that we contact authorities, or that we can send help. KidChat cannot call emergency services or contact anyone outside KidChat. (C-15) This is stated plainly to both parent and child, in calm language, without frightening the child.

Alpha staffing is honest: a Safety Owner (Founder / designated KidChat Safety Lead) reviews the queue. There is no 24/7 operation and no response-time SLA claim. (C-14)

B. ESCALATION MATRIX (T0–T5)

Parent notification rule. Parents are not automatically notified of every safety event. In particular, where a signal suggests the child may be at risk from a parent or caregiver, no automatic notification is sent to the account holder, because that notification could increase the danger. Those cases go to human review. (C-14) The parent's standing ability to read the conversation history is unchanged and is disclosed to the child. (C-11)

No implied action. KidChat never states or implies that it has contacted authorities, emergency services, or any third party unless that action actually occurred.

C. HELPLINE DIRECTORY

Crisis resources come from a server-side, jurisdiction-specific directory that KidChat maintains. The AI model is never permitted to generate a helpline number. If a directory entry is stale or missing for the child's jurisdiction, the flow fails closed: the child is directed to a trusted adult rather than shown an unverified number. (C-13)

Each entry carries: jurisdiction, organization, phone/text/web, purpose, age_applicability, last_verified_at, verified_by. V1 covers Mexico and the United States. [TECHNICAL CONTROL REQUIRED — directory table + fail-closed check]

D. WHAT THE CHILD SEES

Calm, short, never alarming. Child-facing copy is reviewed against this policy before shipping.

• "Your grown-up can see your KidChat conversations." (ages 6–9)
• "Your parent or guardian can see your KidChat conversations." (ages 10–12)
• "I can't call anyone for help. If you need help right now, tell a grown-up you trust."
• "KidChat can be wrong. It's a good place to think, not the last word on things that matter."

E. POLÍTICA DE SEGURIDAD INFANTIL — español (México)

• Respondemos a señales de riesgo con orientación adecuada a la edad y líneas de ayuda verificadas de un directorio que nosotros mantenemos; el modelo de IA nunca inventa un número.
• KidChat no vigila las conversaciones en tiempo real y no puede llamar a servicios de emergencia ni contactar a nadie fuera de KidChat.
• No avisamos automáticamente de cada evento de seguridad. Si la señal sugiere riesgo por parte del padre, la madre o una persona cuidadora, no enviamos aviso automático al titular de la cuenta; el caso pasa a revisión humana.
• Los casos más graves los revisa una persona designada de KidChat. No afirmamos operación 24/7 ni tiempos de respuesta garantizados durante el Alfa.
• Mensajes a la niñez: "Tu adulto puede ver tus conversaciones de KidChat." · "No puedo llamar a nadie por ti. Si necesitas ayuda ahora, dile a un adulto en quien confíes."

F. JURISDICTION DIFFERENCES

[COUNSEL REVIEW REQUIRED — MANDATORY REPORTING POSTURE, MX AND US]

G. BEFORE PUBLIC LAUNCH

Documented coverage, escalation SLA, backup reviewer, reviewer training, incident log, and counsel-approved procedures. Do not launch publicly while T3–T5 depend on an informal process. Alpha operates with invited families only, under this stated limitation.