Capability · 01
Composite scoring
Model + agent + MCP + regulatory dimensions · weighted by deployment context.
What this module does
Per-agent and per-model composite risk grounded in evidence, prompt-injection history, MCP scope abuse history, output-policy violations, drift signals, and EU AI Act risk-tier classification. Briefings translate every finding into business-impact language a board member can act on.
Capabilities
Four operational capabilities.
Capability · 02
Drift detection
Behavior baselines · anomaly signals · production-inference monitoring.
Capability · 03
Cited briefings
AI-generated CISO briefings with cited source events · prompt-injection attempts blocked, MCP scope abuse denied.
Capability · 04
Risk-tier classification
EU AI Act risk-tier classifier applied per agent · Article 13 transparency, Article 9 risk obligations.
Technical detail
What evaluators want to know.
Scoring fabric
Deterministic + LLM reasoning · same architecture as Crypto Posture Management.
Drift signals
Embedding drift · output-distribution drift · invocation-pattern drift.
Briefing generation
Per-agent briefings with cited source events · refresh on-event and weekly.
API access
Programmatic access to scores, briefings, source events.