Regulation (EU) 2024/1689

EU AI Act Compliance

Sentinel compliance mapping across the governance stack — articles actively enforced, tools integrated, and evidence chain documented.

6 / 10
Articles Actively Supported
6 Active 2 Partial 2 Planned
Generate Annex IV Report

Article Mapping

10 articles mapped across Sentinel's governance stack
Article Requirement Sentinel Feature Tool Status
Art. 9 Risk Management System Gate pipeline with composable security gates (PII, injection, tool approval) Sentinel Policy Engine Active
Art. 12 Automatic Record-Keeping Full request lifecycle logging — every gate evaluation, decision, and human action recorded Sentinel Audit Log + Langfuse Tracing Active
Art. 13 Transparency to Deployers Admin dashboard with real-time metrics, gate explanations, and policy documentation Sentinel Admin Console Active
Art. 14 Human Oversight HITL approval queue with approve / reject / override + emergency kill switch (Art. 14(4)(e)) Sentinel Approval Queue Active
Art. 15 Accuracy, Robustness, Cybersecurity Prompt injection detection (LLM Guard), PII detection (Presidio), input / output validation Sentinel Gates + LLM Guard + Presidio Active
Art. 19 Log Retention (6-month min) Configurable retention policy with audit log persistence Sentinel Database Planned
Art. 50 AI Content Transparency Agent metadata for AI-generated content attribution Sentinel Agent Registry Planned
Art. 71 EU Database Registration Agent registry with risk classification, intended purpose, EU registry ID field Sentinel Agent Registry Active
Art. 72 Post-Market Monitoring Metrics pipeline with hourly rollups, Prometheus alerting, anomaly detection Prometheus + Sentinel Metrics Partial
Art. 73 Incident Reporting Webhook notifications for security events with severity classification Sentinel Webhooks Partial

Evidence Chain

Request lifecycle through the compliance pipeline
Agent Registered
Art. 71
Risk Classified
Art. 9
Policy Assigned
Art. 9
Request Arrives
Art. 12
Gates Evaluated
Art. 12 Art. 15
Decision Made
Art. 9
If REVIEW: Human Notified
Art. 14
Logged
Art. 12
Metrics Aggregated
Art. 72
Webhooks Fired
Art. 73
Logs Retained
Art. 19

Cross-Framework Alignment

EU AI Act mapped to NIST AI RMF and ISO 42001
EU AI Act NIST AI RMF ISO 42001 Sentinel Feature
Art. 9 GOVERN 1.4–1.5 8.2–8.3 Gate pipelines per risk level
Art. 12 MAP 1.5, MEASURE 2.6 A.6.2.6 Enriched audit log + metrics rollups
Art. 13 GOVERN 1.7, MAP 5.1 A.8.3 Admin dashboard + A2A Agent Cards
Art. 14 GOVERN 1.3 A.8.5 HITL approval queue + kill switch
Art. 15 MEASURE 2.5–2.7 A.6.2.4 Gate accuracy tracking + Prometheus
Art. 72 MANAGE 4.1 9.1 Metrics pipeline + webhooks

Integrated Tools

Components in the Sentinel stack and their compliance roles

Sentinel Proxy
Core Gateway

Policy enforcement, gate pipeline, decision logging at the API boundary

Art. 9 Art. 12 Art. 14 Art. 15
Langfuse
Observability

OpenTelemetry tracing with full request lifecycle observability and cost attribution

Art. 12
LiteLLM
Provider Router

Multi-provider routing, model allowlisting, and cost tracking across LLM providers

Art. 15
Prometheus
Metrics

Metrics collection, alerting rules, and SLA monitoring for post-market surveillance

Art. 72
Presidio
PII Detection

PII entity detection across 17+ entity types — PERSON, EMAIL, PHONE, IBAN, and more

Art. 15
LLM Guard
Injection Detection

Prompt injection and jailbreak detection with configurable sensitivity thresholds

Art. 15