EU AI Act high-risk classification — credit scoring
You are building a credit-scoring AI for three EU markets. This run resolves the EU AI Act's harmonised high-risk classification from fetched article text — Articles 6 and 27 in full, the Annex III entry via search — while the DE/FR/IT overlay sweeps returned adjacent material only (a BfDI statement on Germany's implementing act, Italy's Legge 132/2025) and are marked as capture gaps.
How the gateway answered
Captured 2 July 2026 — a point-in-time run; corpora and tools have been updated since.
Compare AI Act high-risk classification for credit scoring across DE, FR, and IT. Include the harmonised EU position (Regulation (EU) 2024/1689 Annex III) plus any national supervisory-authority guidance. Side-by-side, with citations.
search(query="high-risk AI system creditworthiness credit score eva…" jurisdictions=[EU])→ 6 results from 18 servers — 2 on-point: the AI Act Annex III creditworthiness row (point 5(b), with eur-lex anchor) and the AI Office prohibited-practices guidelines; 4 off-topic rows from other EU corpora (batteries, roadworthiness, vehicle regulations) not usedget_provision(law="ai-act" article="art_6" jurisdiction="EU")→ Full text of Article 6 — classification rules, the Art. 6(3) derogation and its profiling override, the Art. 6(4) documentation + registration duty, and the Commission-guidelines deadline of 2 February 2026get_provision(law="ai-act" article="art_27" jurisdiction="EU")→ Full text of Article 27 — FRIA duty naming deployers of Annex III points 5(b) and (c) systems, the six assessment elements, market-surveillance notification, and the GDPR Art. 35 DPIA complementget_provision(law="ai-act" article="annex_III" jurisdiction="EU")→ FAILED — 'No provision matches EU AI_ACT annex_III.' The annex is not directly addressable by this ref; the point 5(b) text was resolved via the search row insteadsearch(query="Kreditwürdigkeit KI-System Hochrisiko künstliche Inte…" jurisdictions=[DE])→ 5 results from 7 servers — none on-point for credit-scoring AI; nearest is a BfDI statement on the German AI Act implementing act (KI-VO-DG); the rest are patent-court decisions, a curriculum regulation, and BetrVG § 95search(query="intelligence artificielle haut risque évaluation solv…" jurisdictions=[FR])→ 5 results from 5 servers — consumer-credit creditworthiness provisions (Code de la consommation L313-18) and Conseil constitutionnel credit decisions; no AI Act positionsShow all 8 tool calls
search(query="intelligenza artificiale alto rischio affidabilità cr…" jurisdictions=[IT])→ 5 results from 5 servers — Legge 132/2025 art. 3 (Italy's national AI law, GPAI-scoped) plus off-topic constitutional cases on regional banking powers; no credit-scoring AI positionsearch_guidance(query="AI Act high-risk classification credit scoring credit…" jurisdictions=[EU])→ 2 results from 3 servers — AI Office prohibited-practices guidelines (Art. 5(1)(c) social scoring, mentioning credit-scoring/creditworthiness) on-point; UN R155 vehicle-cybersecurity hit off-topicEU AI Act high-risk classification — credit-scoring AI for DE, FR, IT
The EU position below comes from fetched text: Article 6 and Article 27 of Regulation (EU) 2024/1689 were retrieved in full, and the Annex III credit-scoring entry arrived through search after a direct annex lookup found no provision. The three national-overlay sweeps ran and are reported exactly as they came back — adjacent material, no credit-scoring positions.
The classification test — Article 6 (fetched in full)
Article 6 sets two routes into high-risk status: the Annex I product-safety route (Art. 6(1)) and the listed-use-case route:
"In addition to the high-risk AI systems referred to in paragraph 1, AI systems referred to in Annex III shall be considered to be high-risk." — Art. 6(2)
The Art. 6(3) derogation removes Annex III systems that do "not pose a significant risk of harm to the health, safety or fundamental rights of natural persons" and meet one of four conditions (narrow procedural task; improving a previously completed human activity; pattern-detection that does not replace human assessment; a preparatory task). Its final sentence closes that exit for most scoring systems:
"Notwithstanding the first subparagraph, an AI system referred to …
Every claim traces to a source you can open
A typical AI assistant invents a citation that looks plausible. Ansvar retrieves the real one. Every finding above rests on one of these 7 sources — official legislation, and guidance from standards bodies or regulators — linked wherever the source is publicly reachable; catalog rows served through the gateway are quoted as fetched. No citation is fabricated — every source was retrieved through Ansvar and can be checked.
- AI Act Article 6 — Classification rules for high-risk AI systemsEU · regulation · eur-lex.europa.eu
- AI Act Article 27 — Fundamental rights impact assessment for high-risk AI systemsEU · regulation · eur-lex.europa.eu
- AI Act Annex III, point 5(b) — creditworthiness / credit score (resolved via search; direct annex lookup found no provision)EU · regulation · eur-lex.europa.eu
- Commission (AI Office) Guidelines on prohibited AI practices — Article 5(1)(c) social scoring sectionEU · guidance · digital-strategy.ec.europa.eu
- BfDI Stellungnahme — hearing on the German AI Act implementing act (KI-VO-DG); adjacent, not a credit-scoring positionDE · guidance · bfdi.bund.de
- Code de la consommation, art. L313-18 — creditworthiness assessment in consumer credit (no AI dimension)FR · regulation · legifrance.gouv.fr
- Legge 23 settembre 2025, n. 132, art. 3 — Italy's national AI law (fetched provision is GPAI-scoped, not credit scoring)IT · regulation · normattiva.it
Run this on your own data
This is available as a service: AI Act Readiness Assessment. Bring your own documents and scope, and we'll run it end-to-end — every finding cited and validated by the expert who delivers it.
Read the full sample deliverable: Sample AI Act readiness assessment.
See the coverage behind this run on the AI governance sector page.