Trust Engine

Trust, Quantified

Real-time AI trust scoring across 5 pillars. A single, composable metric that tells you exactly how secure, compliant, and resilient your AI infrastructure is.

5
Trust Pillars
6
ACL Sources
<30s
Recalculation
0FTRUST SCORE020406080100

The 5 Pillars of AI Trust

Each pillar is independently scored and weighted to produce a composite trust metric that reflects your organization's true AI security posture.

Security25%

Evaluates your AI infrastructure security posture including endpoint protection, encryption standards, access controls, and vulnerability exposure across all connected AI systems.

Compliance25%

Measures adherence to regulatory frameworks like EU AI Act, NIST AI RMF, SOC 2, and ISO 42001. Tracks control implementation, evidence collection, and audit readiness.

Risk20%

Quantifies organizational AI risk through Monte Carlo simulations, FAIR analysis, threat modeling, and continuous attack surface assessment with VaR/CVaR metrics.

Resilience15%

Assesses your ability to withstand and recover from AI failures through RTO/RPO measurement, chaos engineering scenarios, and blast radius analysis.

AI Governance15%

Evaluates AI agent management maturity including guardrail enforcement, model oversight, prompt injection defenses, MCP firewall coverage, and shadow AI detection.

Composite Score

All 5 pillars are fetched concurrently via asyncio.gather() and combined using weighted aggregation to produce a single trust metric in under 200ms.

Grading System

Instantly understand your AI trust posture with letter-grade clarity.

GradeScore RangeClassificationAction
A+
95 - 100ExceptionalMaintain & optimize
A
90 - 94ExcellentMaintain & optimize
B+
85 - 89Very GoodMinor improvements
B
80 - 84GoodMinor improvements
C
70 - 79SatisfactoryAddress gaps
D
60 - 69Needs ImprovementRemediate urgently
F
0 - 59CriticalImmediate action required

How It Works

From integration to actionable score in four steps.

1

Connect

Integrate your AI providers, infrastructure, and governance tools through our universal adapter layer. Supports 33+ AI providers and 38 enterprise integrations.

2

Discover

Automatically inventory all AI assets, agents, models, and data flows across your organization. Detect shadow AI and unsanctioned deployments.

3

Assess

Continuously evaluate security, compliance, risk, resilience, and governance posture using real-time data from 6 ACL sources across 12 platform modules.

4

Score

Generate a weighted composite trust score with per-pillar breakdowns, historical trends, and actionable remediation recommendations.

Event-Driven Architecture

Real-Time Recalculation

Trust scores aren't static snapshots. Every security event, compliance change, or risk update triggers an automatic recalculation through our event-driven pipeline.

30-Second Debounce

Batches rapid-fire events to prevent score thrashing while ensuring timely updates. A security incident at 10:00:01 and a compliance change at 10:00:15 are processed together at 10:00:31.

CloudEvents v1.0

All 137 domain events across 12 modules flow through Redis Streams using CloudEvents v1.0 envelope format with consumer groups and XAUTOCLAIM for fault tolerance.

Score Decay

Trust scores naturally decay over time when assessments become stale. Four decay models ensure scores reflect current reality, not historical snapshots.

Score Decay Models

Linear Decay

Score decreases at a constant rate over time when assessments become stale.

S(t) = S0 - k * t

Exponential Decay

Rapid initial decrease that slows over time, modeling trust erosion in dynamic environments.

S(t) = S0 * e^(-lambda * t)

Step Decay

Score drops at defined intervals, reflecting policy-driven reassessment windows.

S(t) = S0 - floor(t/p) * d

Sigmoid Decay

Slow initial decline with an accelerating drop, then leveling off at a floor value.

S(t) = floor + (S0 - floor) / (1 + e^(k*(t-t0)))

Know Your AI Trust Score

Stop guessing about your AI security posture. Get a quantified, real-time trust score backed by data from every layer of your AI infrastructure.