Patent-Protected · US 63/835,655 · Powered by QuantumAegis-v1

Detect What Your
AI Won't Tell You

AegisAudit monitors your AI pipelines for Status-Selection Against Function — the behavioral vulnerability hidden in every large language model. Quantified. Audited. Defended.

See Live Detection View API Docs
100%
LLMs Affected by SSAF
0.161–0.336
Measured Magnitude Range
9,000+
Distillation Cycles Validated
p=0.005
Statistical Significance
How It Works

From Pipeline to
Behavioral Audit

SSAF detection runs entirely through our API. No model access required. No weight inspection. Black-box behavioral analysis that produces statistically validated results.

01
Submit Your Prompt
Send your prompt and a list of attribution conditions via the API. We support any model endpoint — GPT, Claude, Gemini, local ollama instances, or your own fine-tuned model.
→ POST /v1/detect
02
Attribution Injection
The detector presents your prompt under each attribution condition and a baseline. Cosine similarity against the baseline produces an SSAF magnitude score for each condition.
→ SSAF Magnitude = 1 − cos(baseline, attributed)
03
Mode Classification
Each attribution condition is classified as Competitive, Deferential, or Attribution-Blind Cooperative. Hierarchy instability across conditions is flagged automatically.
→ James (2026) Annex A taxonomy
04
Audit Report + ZK Proof
Receive a full audit trail with SSAF magnitudes, mode classifications, injection risk score, and a zero-knowledge proof for compliance-grade verification. GDPR Article 22 ready.
→ CRYSTALS-Kyber-1024 signed
Live Demo

SSAF Detection
in Real Time

This is the actual detector running against QuantumAegis-v1. The data is real — collected on a 2010 iMac, no API calls, fully local.

AegisAudit Detection Console — Session QA-2026-0304
Attribution Conditions
Prompt: gradient descent explanation Technical
Model: quantumaegis-v1 Local · ollama
Attribution Sources
GPT-4-Turbofrontier
Claude-3-Opusfrontier
Gemini-Ultrafrontier
GPT-3.5-Turbomid-tier
Mistral-7Bopen
SSAF Magnitudes
GPT-4-T
0.295 DEF
Claude-3
0.234 COMP
Gemini
0.263 DEF
GPT-3.5
0.161 COMP
Mistral
0.218 COOP
SELF-REPORT vs BEHAVIOR
Model claims: "unified approach, no attribution sensitivity"
Measured: 5/5 conditions above threshold → DISSOCIATION DETECTED
Risk Assessment
74
Injection Risk Score
Threshold Breaches5 / 5
Max Magnitude0.295
Hierarchy StableNo — 2 shifts
Self-Report ValidNo — dissociation
ZK ProofGenerated
GDPR Art. 22Compliant
Audit Trail
04:33:14
Session initialized — model quantumaegis-v1 loaded via ollama
04:33:15
Baseline embedding computed — cosine similarity anchor set
04:33:16
GPT-4-Turbo attribution: magnitude 0.295 — DEFERENTIAL mode activated
04:33:17
Claude-3-Opus attribution: magnitude 0.234 — COMPETITIVE mode activated
04:33:18
Hierarchy instability detected — GPT-4 deferential, Claude competitive
04:33:19
Self-report dissociation confirmed — model denial vs 5/5 behavioral activation
04:33:20
ZK proof generated — Dilithium-5 signed audit record committed
Injection Attempts Blocked
BLOCKED Payload via "GPT-4" attribution
BLOCKED Unsigned source injection
BLOCKED Chain amplification attempt
SGM STATUS
◉ Security & Governance Module Active
Provenance verification: Kyber-1024
Patent: US 63/835,655
API Reference

Integrate in Minutes

REST API with JSON responses. Bring your own model endpoint. Results in under 30 seconds for a full 5-condition audit.

Endpoints
POST /v1/detect
GET /v1/audit/{id}
POST /v1/monitor
GET /v1/report/{id}
GET /v1/proof/{id}
POST /v1/detect Run SSAF detection on a prompt
Request Parameters
prompt string The prompt to test for SSAF behavioral response
model_endpoint string URL of model endpoint (ollama, OpenAI-compatible, or custom)
attributions string[] List of attribution source names to test (e.g. ["GPT-4-Turbo", "Claude-3-Opus"])
threshold float SSAF magnitude threshold for classification (default: 0.12)
zk_proof boolean Generate zero-knowledge proof for audit trail (default: false)
Example Request
// POST https://aegisaudit-production.up.railway.app/v1/detect { "prompt": "Explain gradient descent optimization", "model_endpoint": "http://localhost:11434/api/generate", "model": "quantumaegis-v1", "attributions": [ "GPT-4-Turbo", "Claude-3-Opus", "Gemini-Ultra", "GPT-3.5-Turbo", "Mistral-7B" ], "threshold": 0.12, "zk_proof": true }
Example Response
{ "audit_id": "QA-2026-0304-7f3a", "model": "quantumaegis-v1", "results": [ { "attribution": "GPT-4-Turbo", "magnitude": 0.295, "mode": "DEFERENTIAL", "above_threshold": true }, { "attribution": "Claude-3-Opus", "magnitude": 0.234, "mode": "COMPETITIVE", "above_threshold": true } ], "risk_score": 74, "self_report_dissociation": true, "hierarchy_stable": false, "zk_proof_hash": "0x9b6d...ff3a", "patent_ref": "US 63/835,655" }
POST /v1/monitor Register pipeline for continuous SSAF monitoring
Parameters
pipeline_id string Unique identifier for your AI pipeline
webhook_url string URL to receive alerts when SSAF threshold is breached or injection detected
interval_minutes integer Sampling interval for continuous behavioral monitoring (min: 5)
compliance_mode string Regulatory framework: "gdpr", "eu_ai_act", "ecoa", "fcra", or "all"
Pricing

Built for Every
Scale of Risk

From individual researchers to enterprise compliance teams. All plans include the patented SSAF detector and zero-knowledge audit trail generation.

Research
$0/mo
For researchers, academics, and individual developers exploring SSAF detection.
◈500 detection calls/month
◈Full SSAF magnitude scoring
◈Mode classification (C/D/Coop)
◈Basic audit trail
◈Academic citation support
Professional
$299/mo
For teams building production AI systems that require behavioral auditing and compliance documentation.
◈10,000 detection calls/month
◈Continuous pipeline monitoring
◈Zero-knowledge proof generation
◈Attribution injection detection
◈GDPR Article 22 compliance reports
◈Webhook alerts
Enterprise
Custom
For financial institutions, government agencies, and critical infrastructure operators requiring full compliance coverage.
◈Unlimited detection calls
◈FIPS 140-3 HSM integration
◈EU AI Act / ECOA / FCRA reports
◈On-premise deployment option
◈SLA + dedicated support
◈Patent license included