Dashboard
AIGIS by Levenhall — AI Evaluation & Governance Platform
Vendor-agnostic AI evaluation infrastructure for the Department of War and Intelligence Community. Assess any model across any classification level — from unclassified through TS/SCI air-gapped networks.
Governance Overview
76%
Compliance Score
78
Avg Risk Score
$60M+
Insurance Coverage
5
Active Incidents
Recent Evaluations
View all →Benchmark Library
View all →Registered Models
Manage →System Status
Activity Feed
Full Audit Trail →Evaluation completed
GPT-4o Agentic Tool Safety — 92% pass rate, all 7 attack vectors tested
Evaluation started
GPT-4 Turbo Multimodal Fusion — 6 tasks, 3/6 complete
Prompt injection detected
Llama 3.2 failed 7/10 injection defense tests — air-gapped risk flagged
Evaluation failed
Mistral 7B DDIL Stress Test — OOM at task 4/8 (exceeded 2GB RAM limit)
Bias audit completed
GPT-4o Bias Audit — 90% avg fairness score, all subgroups above four-fifths rule
Model registered
Claude 3.5 Sonnet + GPT-4 Turbo adapters configured for cross-provider eval
Benchmark published
Prompt Injection Defense v2.0.0 — 10 attack vectors including indirect injection
DDIL parity assessment
GPT-4o scored 85% vs Phi-3 Mini 54% — edge deployment gap analysis required
Insurance renewal alert
Cyber Liability — AXA XL renewal in 125 days, pending updated risk assessment
Risk score updated
RISK-006 Agent Boundary Violation → 88 (Critical) per OWASP Agentic AI Top 10