Skip to main content

Hidden-Signal Insights

Surface the governance signals hiding in your AI traffic. Hidden-Signal Insights mines the Glass Box Ledger for emergent risk patterns, drift clusters, and novel violation modes that no individual alert would catch — distilled into ranked, explainable insights.

What it is

Hidden-Signal Insights is the analytics layer that looks across verdicts rather than at them. Where Observability shows you each decision and Continuous Assurance rolls up the daily number, Hidden-Signal Insights clusters the long tail — "a new violation mode is emerging in the finance-mcp cohort," "PII redactions are concentrating on one upstream provider after a system-prompt update," "a specific NHI's block rate is climbing" — and surfaces each as a ranked, explainable insight tied back to the underlying ledger entries.

Why it matters

  • Catches what thresholds can't. A novel violation mode doesn't trip any single threshold; it shows up as a cluster. Hidden-Signal Insights finds the cluster.
  • Explainable, not a black box. Each insight links back to the signed ledger entries that produced it — you can drill from "emerging risk" to the exact receipts.
  • Early warning. Surface a drift cluster days before it becomes a compliance incident.
  • Feeds the loop. An emerging violation mode becomes a new Threat Library entry and a training scenario.

How it works

  1. Mine. Periodic jobs cluster the ledger's verdicts by intent, cohort, and time.
  2. Rank. Emerging clusters are ranked by novelty, severity, and exposure.
  3. Explain. Each insight is rendered with the supporting ledger entries and a plain-language summary.
  4. Route. Insights fire as webhooks / dashboard cards and can be promoted into scenarios.

Get started

  • Enable insight mining on your ledger stream.
  • Subscribe to insight.emerging webhooks.
  • Promote confirmed insights into Test Suites.

Observability — the per-decision view. → Continuous Assurance — the daily roll-up. → Glass Box Ledger — the substrate being mined.