{% extends "base.html" %} {% block title %}AI Trace Auditor — Deterministic AI Compliance Auditing{% endblock %} {% block nav_home %}active{% endblock %} {% block content %}
Open Source Compliance Tooling

Deterministic AI
Compliance Auditing

Audit LLM traces against EU AI Act, NIST AI RMF, ISO 42001, and SOC 2. Every requirement verified against primary legal text with exact clause citations.

{{ total_requirements }}
Total Requirements
{{ regulations|length }}
Regulations
3
Compliance Tiers
5
Trace Formats

Supported Regulations

{% for reg in regulations %}

{{ reg.name }}

{{ reg.nature }}
{{ reg.count }}

{{ reg.description }}

{% if reg.mandatory > 0 %} {{ reg.mandatory }} mandatory {% endif %} {% if reg.recommended > 0 %} {{ reg.recommended }} recommended {% endif %} {% if reg.best_practice > 0 %} {{ reg.best_practice }} best practice {% endif %}
{% endfor %}

How It Works

01

Ingest Traces

Upload OTel, Langfuse, Claude Code, or raw API traces. Multi-agent DAGs detected automatically.

02

Deterministic Scan

Each trace field is checked against verified legal requirements. No LLM guessing; pattern matching against exact statutory text.

03

Tiered Report

Get separate scores for Legal Compliance, Structural Evidence, and Quality. No misleading single number.

Three-Tier Scoring Model

Tier 1: Legal Compliance

Binary pass/fail on law-prescribed checks. Art 12(3) biometric fields, Art 19 retention periods, Art 50 disclosure.

Tier 2: Structural Evidence

Coverage percentage for capability-level requirements. General logging, NIST monitoring, system documentation.

Tier 3: Quality

Best practice observability scores. Token counting, trace linkage, evaluation capture.

{% endblock %}