Regulation (EU) 2024/1689

EU AI Act module

Classify, screen, document and audit AI systems against the EU AI Act. Every verdict is computed by a deterministic Prolog engine — auditable proofs, not checklists.

The AI Act module reuses the same formal-verification engine as the FDA tools: your answers are compiled into Prolog facts and checked against the regulation's rules, so each classification and obligation is provably derived — not an LLM opinion.

The workflow

Register a system

Add your AI system to the registry with its role (provider, deployer, importer, distributor).

Classify risk

Answer the screening questions; the engine returns prohibited, high-risk, limited, minimal or GPAI.

Document & audit

Build the Annex IV technical file and run an article-by-article audit with formal validation.

Assess

System registry

Inventory your AI systems and track each one's classification, obligations and validation status.

Risk classification

A Prolog-backed wizard that decides the risk tier in priority order.

  • Article 3 scope check
  • Article 5 prohibited practices (terminal)
  • GPAI / systemic-risk models
  • Article 6 & Annex III high-risk uses
  • Article 50 limited / minimal

Screen & document

Article 5 screening

Check a system against the eight prohibited AI practices — a prohibited verdict is terminal.

Article 50 transparency

Disclosure and synthetic-content labelling obligations for limited-risk systems.

GPAI obligations

General-purpose AI: transparency, copyright policy and training-data summary (Art. 53), plus systemic-risk duties.

Annex IV technical file

Build the nine-section technical documentation required for high-risk systems.

Operate

Full AI Act audit

An article-by-article audit that validates each provision and produces an evidence report.

Cross-regulation

See where FDA 21 CFR and the EU AI Act overlap (e.g. SaMD ↔ high-risk, Part 11 ↔ data governance).

How classification is decided

The engine evaluates outcomes in strict priority order and returns the first that matches:

  1. Prohibited — the system meets an Article 5 practice (terminal).
  2. GPAI / systemic — a general-purpose model, flagged systemic above the compute threshold.
  3. High-risk — an Article 6 safety component or a listed Annex III use (unless Art. 6(3) exempt).
  4. Limited — Article 50 transparency obligations apply.
  5. Minimal — no specific obligations beyond voluntary codes.

Because the decision is a Prolog program, the basis for every verdict is recorded as a list of the exact rules that fired — ready to export as evidence.

Start now