Live Classification Engine

Classify your AI system under the EU AI Act

Describe your AI system in plain language — in any of 6 languages. Our multilingual LLM extracts the key facts. Then our deterministic engine classifies risk against Articles 5, Annex III, GPAI, and national laws. LLM extracts facts. Deterministic rules determine the result.

REGULATION (EU) 2024/1689Free · No signup · Results in seconds

azcomply-engine current

ACTIVE

ENEU AI Act obligations only — no country-specific national law

Ctrl+Enter to classify0 / 2,000 words

> select a scenario or describe your AI system above█

EU 2024/1689 · stateless · no data retained

64 articles mapped

LLM extraction · 6 languages

Deterministic classification

6 national law overlays

64 Articles Mapped

Full Assessment Report

What you get beyond the free questionnaire

The free engine gives you a verdict. The full report gives you a compliance roadmap — 9 AI advisors analyse your system across 8 report sections.

Free Risk Preview€0 · zero credits

Risk classification level
Key obligations count
Applicable articles
Full obligations checklist
90-day action plan
FRIA analysis
Vendor gap assessment
National law overlays

Recommended

Full ReportFrom €0.20/credit

Risk classification level
Key obligations count
Applicable articles
Full obligations checklist
90-day action plan
FRIA analysis
Vendor gap assessment
National law overlays

Get Full Report

1 credit per assessment · PDF download

Architecture

How grounded classification works

FreeText → Engine → JSON · No LLM · <100ms

Text description

Deterministic Engine13-layer pipeline · <100ms

Risk + Articles + Fines

Full ReportDocument → Extract → Engine → Narrative → PDF

ExtractText / PDF / Doc / Image

Deterministic Engine

Source of truth

AI as WriterGrounded narrative

PDF Report

The AI can never override the Engine

If the engine says HIGH RISK, the AI is legally bound to that classification. No hallucination. No downgrading. The engine is the judge — the AI is the writer.

Non-Compliance Risk

The cost of not knowing your risk level

Maximum fine

€35M

or 7% of global annual turnover

Full compliance report

€0.20

one-time · 12-page PDF · 6 languages

Get Your Report Now

Comparison

Rule-based vs. LLM-based classification

AZComply Engine

Rule-based · Deterministic

Engine type

Deterministic rules

Response time

<100ms

Citations

Article-level

Cost per call

€0 (free tier)

Reproducibility

100% deterministic

Data residency

EU only (NL/DE)

National laws

6 country overlays

Generic AI

LLM-based · Variable

Engine type

LLM-based (GPT)

Response time

5-15 seconds

Citations

Generic advice

Cost per call

€0.03-0.10

Reproducibility

Variable outputs

Data residency

US servers

National laws

None

FOR DEVELOPERS & INTEGRATORS

Classification Engine API

Integrate EU AI Act risk classification directly into your compliance pipeline. Same deterministic engine — zero LLM cost, sub-second response.

Your Integration

One API call between your AI and legal certainty

Your AI handles the conversation. Our engine handles the law. Drop one API call into your pipeline — your AI extracts, the engine classifies, your AI responds grounded.

User input

From your app

Your AI extracts

Context & facts

AZComply Engine

Deterministic

13-layer pipeline · <100ms · EU-hosted

Your AI responds

Grounded by engine

Safe output

Legally grounded

Your stack

Our engine

Your stack

No retraining needed

The engine is deterministic — same input, same classification, every time. Auditable and legally defensible.

One line of code

POST to /v1/classify. Get risk level, article citations, obligations, and fine exposure back in 47ms.

REST API with JSON response

Deterministic classification — no LLM hallucination

National law overlays (BE, NL, FR, DE, IT, ES)

Webhook notifications on classification changes

CustomOrganization entitlement required

Contact sales View API Docs

api.azcomply.eu

RequestPOST /v1/classify

curl -X POST https://api.azcomply.eu/v1/classify \
  -H "Authorization: Bearer ak_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "system_name": "CV Screening Tool",
    "description": "AI tool that ranks job applicants...",
    "jurisdictions": ["BE", "NL"],
    "operator_role": "DEPLOYER"
  }'

Response200 OK · 47ms

{
  "risk_level": "HIGH_RISK",
  "articles": ["Annex III §4(a)", "Art. 26"],
  "obligations": ["Art. 9", "Art. 11", "Art. 13", "Art. 14"],
  "fria_required": true,
  "compliance_deadline": "2027-12-02",
  "fine_exposure": "€15M or 3% turnover",
  "engine_version": "current"
}

FAQ

Frequently asked questions

How does the classification engine work?+

The engine uses a deterministic 13-layer pipeline written in Python. It checks your AI system description against Article 5 prohibited practices, Annex III high-risk categories, GPAI thresholds, and national law overlays. No large language model is used in the classification path, so results are 100% reproducible.

Is the free questionnaire really free?+

Yes. The free questionnaire is stateless: no account required, no saved workspace, no credits deducted. It gives a preview risk result and article signals; logged-in classification and paid assessment use the full QAE pipeline.

How accurate is the classification?+

The engine maps directly to EU Regulation 2024/1689 text. Every classification cites specific articles, annexes, and recitals. Because it's rule-based (not LLM-generated), results are deterministic — the same input always produces the same output.

What jurisdictions are supported?+

EU-wide classification under the AI Act, plus national law overlays for Belgium (CAO 9/39/81/100), Netherlands, France, Germany, Italy, and Spain. Each overlay adds country-specific obligations from labour law, data protection, and sector regulation.

Where is my data processed?+

All processing happens on EU servers in the Netherlands (europe-west4). Database storage is in Frankfurt (eu-central-1). No data ever leaves the European Union. The free endpoint doesn't store any data at all.

What's the difference between the free questionnaire and the full assessment?+

The free questionnaire gives a preview risk result and article signals with no account. Logged-in classification uses QAE extraction and the deterministic engine; the full assessment adds the advisor pipeline, narrative report, action plan, FRIA requirements, and downloadable PDF.

Enforcement begins December 2, 2027

Ready to classify your AI system?

Start with the free questionnaire, or sign in to run QAE classification and upgrade to a full compliance report with an action plan.

Full Assessment Try Free Questionnaire

No account needed · GDPR-compliant · EU-hosted · Data never stored

Know your risk level

Free · No signup · <1s

Classify

EU AI Act Classification Engine | AZComply