Is AI safe for K-12 students?

AI used with children is only as safe as its guardrails. The good news: responsible tools exist. The bad news: many consumer chatbots were never designed for kids. Here's what separates safe classroom AI from a risk.

The non-negotiables

• Adult in the loop — a teacher or parent supervises every session; no autonomous contact with children.
• Hardcoded content guardrails — non-overridable blocks on harmful content.
• Age-appropriate language — vocabulary and complexity matched to the grade.
• Data privacy — minimal data collection, secure storage, and clear policies.
• Risk flags — the system alerts an adult to signs of distress or safeguarding concerns.

Questions to ask any AI tool

• Can a child use this without an adult present? (It shouldn't be possible.)
• What data is collected, and where is it stored?
• Can the safety rules be turned off by configuration? (They shouldn't be.)
• How are crisis or safeguarding signals handled?

Why supervision is the foundation

No content filter is perfect, which is why the single most important safeguard is a human adult in control. Safe classroom AI is designed to assist a teacher, never to replace them — and never to operate on its own.

How Saya is built

Saya has no autonomous mode: it can't initiate contact with students, can't run without a teacher present, and enforces hardcoded safety rules that no setting can override — plus private real-time risk flags to the teacher. Safety is the foundation, not a feature.