KeepChatSafe is a governance gateway wrapped in a chat students want to use. Every capability below ships as part of one product — organized around the four commitments a school makes when it puts AI in front of students.
The safety layer runs on every message, in both directions — before anything reaches an AI model and again before anything reaches a student.
Purpose-built safety classifiers screen for self-harm, sexual content, eating disorders, dangerous activities, and romantic roleplay — categories defined for adolescents, built on open safety models published by the frontier AI labs themselves.
A serious disclosure triggers a supportive response to the student and an immediate, context-rich alert to the counselor or administrator your district designates. Freezing a student mid-disclosure is a failure mode we designed out.
Names, student IDs, contact details, and other personal information are redacted before any message leaves your governed environment — and responses are screened on the way back.
A dedicated scanner detects prompt-injection and guardrail-bypass attempts before they reach the model, blocks them, and logs them for review — because teenagers are nothing if not creative.
Adoption is a safety feature: a sanctioned tool only protects students if it's good enough that they choose it.
Fast, clean, and distraction-free on managed Chromebooks and district devices. One conversation surface, no feature maze, no feed.
Teachers set the mode per class: Socratic guide-don't-answer tutoring for homework, open assistance for research projects. Academic integrity becomes a policy your educators operate — not a hope.
Teachers see their class's activity, review flagged conversations, and adjust class-level settings without filing an IT ticket.
Your district operates the controls. KeepChatSafe provides the instrument, the enforcement, and the evidence.
Your wellbeing and safety team defines the content criteria — categories, thresholds, escalation routes. Every change is recorded as a signed, attested policy decision made by your district.
Not every AI provider permits minors as end users. The routing layer excludes providers whose terms prohibit it and enforces minor-safety screening on the rest — by policy engine, not configuration promise. To our knowledge, no other platform does this.
Every conversation, verdict, escalation, and administrative action lands in your district's isolated tenant. If the audit log cannot be written, the request does not run.
School deployments ship with the ungoverned path removed — absent from the build, not switched off in a config file. There is no flag that routes a student around the safety layer.
Conversation retention and deletion follow your records-management policy. You set the schedule; the platform enforces it and produces the evidence.
Built to slot into district identity, family communication, and procurement — the way K-12 infrastructure is actually bought and run.
Students use the school accounts they already have. KeepChatSafe connects to your existing identity provider — Google Workspace, Microsoft 365, Clever, or ClassLink — with OneRoster rostering support. School identity doubles as age assurance: no self-reported birthdays.
A family-facing surface for program disclosure and consent, designed to align with your district's existing annual-notification workflows.
Control-mapped evidence for FERPA, COPPA, and California student-privacy law (SOPIPA / AB 1584), plus a dedicated pack for AI-provider under-18 requirements — exportable for auditors.
Security questionnaires, data-flow diagrams, DPA templates drafted to AB 1584's required clauses, and a straight-answer security review. We know what you have to ask, because we'd ask it too.
District SSO establishes identity, grade band, and class — age assurance from your rostering, not from self-reported birthdays.
Personal information redacted; safety classifiers and injection defense run before anything leaves your governed environment.
Routed only to providers whose terms permit serving minors, under your district's policy. If governance is unavailable, the request fails closed — nothing is sent raw.
Output screened again; wellbeing flags escalate to your designated staff; every step lands in the tamper-evident audit trail.
Bring your acceptable-use policy, your filtering standards, and your hardest questions. We'll map every capability to them, live.
Book a district briefing