GDPR Compliance Center
LiveReal-time GDPR compliance monitoring across the entire agent fleet. Every metric is sourced from live systems — not simulated.
Fleet GDPR Score
47%
weighted composite · 5/9 verified
EU Data Residency
2
agents EU-only certified
GDPR Articles
7/8
articles actively enforced
Active Alerts
2
PII / policy violations
📋 Agent Compliance Matrix
| Agent | GDPR | AI Act | EU Res. | Trust |
|---|---|---|---|---|
ConvoGuard AI | ✓ | ✓ | ✓ | 95 |
Trust Verifier | ✓ | — | ✓ | 92 |
Semantic Aligner | — | ✓ | — | 88 |
Deadline Enforcer | — | — | — | 90 |
Fairness Auditor | ✓ | ✓ | — | 87 |
Trust Protocol | — | — | — | 91 |
Capability Broker | — | — | — | 94 |
ReelBerlin Engine | ✓ | — | — | 85 |
Quantum Grid Resilience | ✓ | ✓ | — | 96 |
📜 GDPR Article Coverage
ConvoGuard ConsentRule — verifies explicit consent before health data collection
ConsentDetector — flags HIV, mental health, medication data with SIGNAL_GDPR_SPECIAL_CATEGORY
PoE-A2A public claims endpoint (/.well-known/poe-claims.json) — all processing activities published with Ed25519 signatures
ConversationRepository.delete() — GDPR-compliant data deletion port
PendingAction human-in-the-loop workflow — high-risk decisions require explicit human approval
Local-first ONNX inference — no data leaves the processing boundary
PoE back-linked chain — automatic cryptographic log of all processing activities
DeclarationOfConformity entity — auto-generated EU AI Act Article 47 conformity assessment
🚨 PII & Policy Violation Alerts2 ACTIVE
PII detected in agent output logs — ReelBerlin Engine
Agent accessed restricted health data category without consent gate
Email address pattern in Semantic Aligner translation cache
Fairness Auditor bias score drifted 3% — within tolerance
Proof of Execution Chain
Ed25519 Signed · Back-Linked · Solana Anchored
Live GDPR Consent Checker
Powered by ConvoGuard · GDPR Art 6/9 · Real-time Neural Inference
📊 ConvoGuard Accuracy BenchmarksGET /api/gdpr/benchmarks
Crisis Detection
100%
recall · F1: 0.97
n=312
Consent Detection
96%
precision · F1: 0.94
n=847
Art 9 Detection
89%
recall · F1: 0.90
n=234
ONNX Latency
8ms
p50 · p95: 18ms
p99: 42ms
Methodology: Vitest test suite on curated conversation fixtures. Crisis recall prioritized over precision — false positives acceptable, false negatives not.
⚠️ Transparency & Known Limitations
We believe honest disclosure of limitations builds more trust than inflated claims. This section exists because compliance tools that hide their gaps are more dangerous than having no tool at all.
Consent detection is English-only — multilingual support planned for Q3 2026
Art 9 detector covers 12 medical keyword categories; novel terminology needs rule updates
GDPR score is a composite metric reflecting badge + policy coverage, not legal attestation
Blockchain anchoring on Solana devnet — mainnet migration pending security audit
Cannot detect implied consent or consent given outside the monitored conversation
Does not cover: DPO appointment, supervisory authority communication, or training data governance
End-to-End GDPR Compliance Architecture
Integration & Exports
Built in Berlin · GDPR-native from Day 1 · All data processed within EU boundaries