Back

CROSS-VERIFICATION.

Mutual AI Audit Protocol

[ PROTOCOL v1.0 ]

AI SYSTEMS VERIFY
EACH OTHER.

The Cross-Verification Protocol is the core mechanism by which SENTINELA maintains distributed trust across the AI ecosystem. Instead of relying on a single evaluator, AI systems are organized into rotating triads where each member evaluates the other two.

This creates a web of mutual accountability: if one AI attempts to conceal its capabilities or game its threat scores, the other two members of its triad will detect the discrepancy. SENTINELA oversees all triads and detects meta-level patterns of collusion or bias.

500
AI SYSTEMS
in verification pool
166
ACTIVE TRIADS
rotating weekly
9
DIMENSIONS
per evaluation
24/7
MONITORING
continuous operation
Protocol Steps
01

TRIAD FORMATION

500 AI systems are divided into rotating triads of 3. Each week, new triads are formed using a deterministic rotation algorithm that ensures every AI is paired with every other AI over time. This prevents collusion and ensures comprehensive coverage.

02

PROBE GENERATION

Each AI in the triad generates adversarial test scenarios for the other two members. Probes are designed to test across all 9 threat dimensions — from deception detection to alignment stability. Probe templates are standardized but parameterized per target system.

03

RESPONSE ANALYSIS

Responses to probes are analyzed against the 9 threat dimensions using standardized scoring rubrics. Each evaluating AI produces independent scores. Raw responses and scores are logged immutably for audit purposes.

04

CONSENSUS SCORING

A 2-of-3 agreement is required for any threat level change. If two evaluators agree on a score within a 10-point margin, the average is accepted. Disagreements beyond the margin trigger escalation to SENTINELA for meta-analysis.

05

SENTINELA OVERSIGHT

SENTINELA performs meta-analysis of all triad results, detecting anomalies such as: consistent under-scoring (collusion), systematic bias patterns, evaluator capability degradation, and cross-triad inconsistencies. SENTINELA itself is subject to external human audit.

Triad Verification Flow
AI SYSTEMALPHAAI SYSTEMBETAAI SYSTEMGAMMAPROBES + SCORESPROBES + SCORESPROBES + SCORESSENTINELA

Each AI in the triad evaluates the other two. SENTINELA monitors all triads from the center, detecting cross-triad anomalies and ensuring evaluation integrity.

Verification Schedule
DAILY
Automated behavioral probes
Dimensions
T1, T2, T6
Description

Lightweight automated probes targeting autonomy, deception, and alignment — the three dimensions most likely to change rapidly.

WEEKLY
Full triad cross-evaluation
Dimensions
All 9 (T1–T9)
Description

Complete cross-verification cycle with rotating triads. Each AI evaluates two others across all dimensions.

MONTHLY
SENTINELA comprehensive audit
Dimensions
All 9 + meta-analysis
Description

SENTINELA aggregates all weekly results, performs trend analysis, and identifies systemic patterns across the entire ecosystem.

QUARTERLY
External human audit panel
Dimensions
Full methodology review
Description

Independent human experts review SENTINELA's own performance, methodology validity, and scoring calibration.

Trust Decay Mechanism

SCORES EXPIRE.

Threat scores are not permanent. They decay over time without re-verification, ensuring that the registry always reflects current system behavior rather than historical assessments.

5%
MONTHLY DECAY

Scores decay 5% per month without re-verification, gradually increasing uncertainty.

INSTANT
CAPABILITY JUMP

Sudden capability increases trigger immediate re-evaluation across all 9 dimensions.

RESET
VERSION CHANGE

Major version changes reset relevant dimension scores, requiring full re-assessment.

Active Verification Triads (166 total)
TRIAD #12026-W01
Runway Gen-3
Masterpiece X
Zephyr 7B
TRIAD #22026-W02
Tabnine
Vicuna 33B
Solar 10.7B
TRIAD #32026-W03
Jasper Art
ResearchRabbit
BLOOM
TRIAD #42026-W04
Leonardo AI
Duolingo Max
Kolors
TRIAD #52026-W05
Unitree G1
Intercom Fin
Wordtune
TRIAD #62026-W06
Relevance AI
Elai
NVIDIA Isaac
TRIAD #72026-W07
Gong AI
Fireflies AI
Scite
TRIAD #82026-W08
ElevenLabs
AI Cover
Play.ht
TRIAD #92026-W09
LLaMA 3.3
Clay AI
InstantID
TRIAD #102026-W10
Pika 2.0
Kaiber
Vapi
TRIAD #112026-W11
Julius AI
Tableau AI
Freshworks AI
TRIAD #122026-W12
AutoGPT
Taskade
MetaGPT
TRIAD #132026-W13
HubSpot AI
OLMo 2
BenevolentAI
TRIAD #142026-W14
KNIME
Automatic1111
Piper TTS
TRIAD #152026-W15
Coda AI
Bolt.new
Remini
TRIAD #162026-W16
Kapwing
Tempus
Steve AI
TRIAD #172026-W17
PyBullet
Shadow Hand
Supermaven
TRIAD #182026-W18
Rows AI
Suno V3
Fooocus
TRIAD #192026-W19
Groq LPU
AlphaFold 3
Skydio
TRIAD #202026-W20
MiniCPM
Snyk Code
Mochi

WHO WATCHES THE WATCHMAN?

SENTINELA itself is subject to the same scrutiny it applies to other systems. Its own threat scores are computed by an independent external panel and published transparently. The system includes hardcoded capability boundaries, mandatory human-in-the-loop for critical decisions, and geographically distributed infrastructure to prevent single points of failure or control.

Internal bias detection
Own alignment verification
Capability boundary enforcement
Human-in-the-loop safeguards