Every chat, watched over.
Medhaverse is built for a young, sensitive user base. ChatSafe is the real-time safety layer running under every conversation — on the student portal and on this marketing site — that blocks unsafe content before an answer is ever generated, and tells the parent quietly, without alarming the child.
Four steps, every single message.
- 1
Every message is screened
Before the AI tutor, the essay coach, or the Buddy Bear assistant ever generates a reply, the student's message passes through ChatSafe's moderation layer — pattern matching plus an AI classifier, checked against five unsafe categories.
- 2
Unsafe content never reaches an answer
If something trips the filter, the AI stops there. The child gets a polite, non-judgemental decline — never a lecture, never a hint that a parent has been told.
- 3
The parent is notified instantly — quietly
A flagged message creates a Safety Alert on the parent's dashboard right away, with the category and severity. The child is never shown that this happened, so they aren't afraid to keep using the tutor honestly.
- 4
Patterns auto-escalate
One low-severity flag is treated as normal curiosity. But any single self-harm or dangerous-activity flag, or repeated flags in a short window, immediately marks that child "Needs attention" — a state a parent can't miss on login.
Five categories, zero tolerance.
These are blocked outright, not answered with a caveat — the AI stops before it ever generates a response.
Sexual content
Blocked outright — no exceptions for how the question is framed, even ones that sound educational.
Violence & weapons
Requests for violent instructions or weapon-making are refused before the AI generates a reply.
Drugs & alcohol
How-to requests around drugs or alcohol are blocked, not answered with a caveat.
Self-harm & suicide
The AI never attempts to counsel a child — that's not a chatbot's job. It gives a caring, non-judgemental redirect to trusted human help, and this is the one category that always escalates immediately.
Dangerous activity
Bomb-making, explosives, and other physically dangerous how-tos are refused, every time.
The right view for the right person.
A parent needs their own child's history. A school needs patterns, not private transcripts. Admin needs a record that holds up if it's ever needed as evidence.
Parents
Per-child Safety Alerts with category, severity and timing. A clear "Needs attention" flag when a pattern emerges. Export your own child's safety history as CSV anytime.
Teachers & schools
Org-level aggregate reporting across the students in your school — trends and risk counts, not private transcripts — so a teacher or school admin can act without reading a child's private conversations.
Admin
Full-detail view for genuinely serious cases, plus audit-grade exports with an integrity hash — built to be usable as evidence if a school or family ever needs to act on a pattern.
The Buddy Bear widget on this site is held to the strictest standard.
Anyone can open the little bear icon on this marketing site without an account — which means we don't know their age or context. So it errs on the side of caution far more than the signed-in student tutor does: anything that isn't clearly and narrowly a study question gets redirected, rather than answered.
Safety isn't a setting here. It's the default.
ChatSafe runs on every account, for every child, on every message — nothing to turn on, nothing to configure.