Why Sentinel exists
The gap between regulation and code
Modern AI regulations introduce complex technical obligations that are difficult to interpret, verify and document consistently. Engineering teams work with source code and CI/CD pipelines. Legal and compliance teams work with regulatory texts and obligation frameworks.
The gap between these worlds leads to manual assessments, inconsistent interpretations and documentation that is difficult to reproduce or audit. Sentinel was built to bridge that gap through evidence-driven technical analysis.
Engineering team
Compliance team
Sentinel bridges the gap
How Sentinel thinks
From artifacts to alignment
Software Artifacts
code · config · docs · CI/CD
Evidence Extraction
AST parsing · 33+ signal patterns
Finding Classification
severity · confidence · provenance
Article Requirements
22 EU AI Act articles
Alignment Scoring
weighted · deterministic · versioned
Compliance Report
signed · reproducible · Annex IV
Each step is traceable. Every finding links back to the artifact that produced it.
Our philosophy
Five principles that guide every decision
Evidence before opinion
Sentinel evaluates observable technical artifacts — source code, configuration, infrastructure, CI/CD pipelines — rather than questionnaires or subjective declarations.
Explainability
Every finding is traceable to the evidence that produced it. No black-box scoring. Each article score can be decomposed into its contributing signals.
Reproducibility
The same software analyzed under the same methodology version always produces the same result. Scores are deterministic, not probabilistic.
Transparency
Methodology, scoring weights and evidence models are documented and versioned. Changes to the rule engine are communicated before deployment.
Human oversight
Sentinel assists experts. It never replaces legal, engineering or regulatory judgment. Every report is designed to be reviewed and challenged by qualified professionals.
What Sentinel is not
Built around evidence
What counts as evidence
Evidence in Sentinel is any observable technical artifact that can be linked to a regulatory requirement. Each evidence item carries its own provenance and confidence classification.
Full methodologyBeyond the EU AI Act
A framework built to extend
Sentinel's architecture supports multiple regulatory frameworks through independent Rule Packs.
EU AI Act
Regulation (EU) 2024/1689
ISO/IEC 42001
AI Management Systems
NIST AI RMF
AI Risk Management Framework