Home Safety & Monitoring Managing KyberPulse Alerts

Managing KyberPulse Alerts

Last updated on Jul 09, 2026

What is KyberPulse?

KyberPulse is KyberGate's AI-powered student safety monitoring system. It continuously scans Google Docs, Gmail, and web browsing activity to detect concerning content including violence, self-harm, bullying, hate speech, and other safety-related patterns.

How Alerts Work

When KyberPulse detects a potential concern, it creates an alert with:

  • Student name and email — who triggered the alert
  • Severity level — Critical, High, Medium, or Low
  • Category — the type of concern detected
  • Source — where it was found (Google Doc, Gmail, or web browsing)
  • Document context — the document name, type, and a content preview with the matched phrase highlighted

Reviewing Alerts

From the KyberPulse page in your dashboard, click any alert to expand it. You have three actions:

  • Mark Reviewed — acknowledge the alert and clear it from your active feed
  • Escalate — flag for further action by a counselor, administrator, or other staff
  • Not a Concern — mark as a false positive. KyberPulse learns from this and will reduce similar alerts over time

Adaptive Learning

KyberPulse gets smarter the more your team uses it. Here's how it works:

  1. When you mark an alert as Not a Concern, KyberPulse records the phrase and category
  2. After the same phrase+category combination has been dismissed 3 times, KyberPulse automatically suppresses future alerts for that pattern
  3. Auto-suppressed alerts are still logged for audit purposes, but they won't appear in your active feed or trigger notifications

Important: Critical severity alerts (threats of violence, self-harm indicators) are never auto-suppressed, regardless of how many times they've been dismissed. Student safety always comes first.

Examples

  • A student writing about the Holocaust in a history essay triggers a "hate speech" alert. After your team marks it "Not a Concern" a few times, similar alerts for that term in academic context stop appearing.
  • Literary analysis of violence in novels (Shakespeare, Elie Wiesel, etc.) is recognized as educational after consistent feedback.
  • Common curriculum topics that match safety keywords are handled automatically.

Managing Learned Exceptions

On the KyberPulse page, look for the Learned Exceptions section. Here you can:

  • See all patterns KyberPulse has learned from your team's feedback
  • View how many times each pattern was dismissed
  • Re-enable any pattern if you want alerts to resume for it

Each school's learned exceptions are independent — what you dismiss for your organization won't affect other schools using KyberGate.

Severity Levels

  • Critical — Immediate safety risk. Always generates a notification, never auto-suppressed.
  • High — Serious concern requiring prompt review (bullying, explicit content)
  • Medium — Moderate concern (hate speech in academic context, inappropriate language)
  • Low — Informational (flagged keywords in clearly academic contexts)

Tips for Reducing Alert Noise

  • Review alerts consistently — the system learns faster with regular feedback from your team
  • Use Not a Concern liberally for academic content — essays, research papers, and curriculum-related content may trigger alerts initially, but a few dismissals will train KyberPulse for your specific curriculum
  • Check Learned Exceptions periodically — make sure auto-suppressed patterns still make sense for your school
  • Critical alerts deserve attention — these are never suppressed and indicate potential immediate safety concerns