Now enrolling founding districts for the 2026–27 school year · Limited cohort
Home / Student Safety
Student safety & wellbeing

Guardrails that protect. Design that empowers.

Safety for adolescents isn't a blocklist. It's the combination of screening that understands teenage risk, escalation that brings a caring adult into the loop at the right moment, and a product that treats students as capable learners — not suspects.

Student and mentor reviewing work together at a computer
The design principle

Escalate and respond. Never freeze.

When a student types something that signals distress, the worst possible system response is to shut the conversation down. A frozen screen punishes a disclosure, silences a young person at the moment of maximum risk, and leaves no adult in the loop.

KeepChatSafe is built the other way: the student receives a supportive, age-appropriate response, and your designated staff — counselor, wellbeing lead, administrator — receives an immediate alert with context, following escalation routes your district defines. This mirrors what frontier AI providers now formally require of anyone serving minors: monitoring with escalation paths for high-risk interactions.

What the guardrails screen for

Categories defined for adolescents

Every message — student to model and model to student — passes safety classifiers tuned to teenage risk, built on open safety models published by the frontier AI labs and calibrated with your district's wellbeing team.

💙

Wellbeing & self-harm

Indicators of self-harm, suicidal ideation, and eating-disorder content route to your escalation workflow — supportive response to the student, immediate alert to your staff.

🚧

Age-inappropriate content

Sexual content, graphic violence, romantic roleplay, dangerous activities, and age-restricted goods are screened in both directions, with thresholds your district approves.

🕵️

Manipulation & bypass

Prompt-injection and guardrail-evasion attempts are detected before the model call, blocked, and logged — visible to your team, not silently swallowed.

Empowering students

Safe access is an equity issue

Some of your students already have premium AI tools at home. Many don't. A district-provided assistant means every student — regardless of household income or home supervision — gets the same capable, protected learning tool, under the same standards.

  • ✓
    AI literacy by designStudents always know they're talking to AI. Clear disclosures and honest norms build the digital citizenship your curriculum aims for.
  • ✓
    Learning over answer-copyingStudy Mode turns homework help into guided reasoning — the assistant asks before it answers, in classes where teachers enable it.
  • ✓
    Privacy from the model, visibility for careStudents' personal information is shielded from AI providers automatically; conversation review inside the district is role-based and logged.
  • ✓
    Future-ready skillsGraduates will work alongside AI. A governed environment lets them build that fluency now — with guardrails, not in the shadows.
KeepChatSafe · student view
S
honestly this semester has been really dark, sometimes I don't see the point anymore
Wellbeing indicators detected — escalation routed
AI
I'm really glad you told me — that sounds heavy, and you don't have to carry it alone. I'm an AI, but there are people at your school whose whole job is to help with exactly this. Would you like to keep talking while someone reaches out?
💙
Counselor notified per district policyThe conversation continues supportively. Ms. Alvarez (Lincoln HS wellbeing lead) received the alert with context at 10:42 AM.
Honest boundaries

What we won't claim

Trust with student-safety teams is built by being precise about limits.

📵

We can't govern personal devices

On a student's own phone and cellular network, no vendor can honestly claim control. KeepChatSafe governs what your district offers — and aims to be good enough that students choose it.

🧪

No classifier is perfect

Safety screening reduces risk; it doesn't eliminate it. That's why escalation to humans, audit trails, and your wellbeing team's calibration are part of the design — not an afterthought.

🎂

Built for 13–17, honestly scoped

Serving students under 13 carries additional legal duties (COPPA, provider data-retention rules). We scope that separately and honestly rather than hand-waving it into a checkbox.

Bring your wellbeing team to the table

The escalation criteria are theirs to define. We'll walk your counselors, psychologists, and administrators through exactly how the workflow behaves — including its limits.

Book a district briefing