Most organisations haven't answered this question with a documented, legally defensible classification decision. The Annex III Classification Guide gives you the three-step test, worked examples across all eight domains, and a signed screening record you retain as evidence.
If your organisation uses AI in employment screening, healthcare decisions, credit scoring, biometric identification, or any of the eight Annex III domains — and you haven't completed a formal classification — you are carrying undocumented legal risk. The EU AI Act does not require intent. It requires evidence.
Article 6 of Regulation (EU) 2024/1689 defines a structured three-step classification test. Most guides describe the categories in general terms. This guide gives you the test itself — worked through for every Annex III domain, with a completed screening record you fill in, sign, and file.
Classification is the gateway obligation. Every other high-risk AI requirement — risk management, data governance, technical documentation, human oversight — depends on completing this step first.
AI Literacy (Art. 4) and GPAI obligations entered into force. Governance frameworks required for all organisations deploying AI within the EU.
Full high-risk obligations apply to standalone AI systems meeting Annex III criteria. Classification and evidence documentation required.
Full obligations apply to AI embedded in regulated products (machinery, medical devices, toys). Classification documentation required from procurement stage.
Your purchase delivers the guided ebook and the companion Excel register as an immediate download. Everything you need to complete, document, and sign off your Annex III classification.
Eight chapters. Covers the full three-step Article 6 test, all eight Annex III domains with worked examples, the Article 6(3) exclusion in full, and a field-by-field screening record template. Written for DPOs, CIOs, and legal leads — not data scientists.
The companion Excel workbook with five structured worksheets: firm and system identification, three-step classification assessment, HR obligation tracker, and a time-stamped audit log. Ready to populate from the first page.
Every chapter has a defined purpose and a defined output. The guide moves from legal obligation to completed evidence record in eight steps.
Why "unassessed" is not a legal position, the consequences of misclassification, and the peer firm scenario that illustrates the gap.
Article 6(1), 6(2), and 6(3) explained in plain language. The full Arts. 9–15 obligation cascade for high-risk systems, with enforcement deadlines.
AI inventory scoping, the one-row-per-system rule, and the five-step pre-assessment checklist. Establish the perimeter before classification begins.
The full Article 6 decision process in a visual, step-by-step instrument. Area screening → use-case specificity → Art. 6(3) exclusion, with outcome branches.
All eight classification domains with legal text, worked examples, and the boundary conditions. Area 8 (Administration of Justice / Democratic Processes) given special depth as the most frequently misclassified category.
The three grounds for exclusion from high-risk classification. Documentation requirements. The critical Art. 50(4) conflation warning.
A 30-day action plan across five steps: obligation scoping, FRIA determination, documentation register setup, human oversight design, and governance notification.
Field-by-field walkthrough of the screening record template: Document Control, System Identification, Three-Step Assessment, and Sign-off. Plus a worked example for GPT-4o in an HR context.
This is not a developer guide. It is a governance instrument for the professional who holds accountability for AI risk in a regulated organisation.
DPOs who need to determine if an AI system triggers Art. 27 FRIA obligations and document the classification basis under Art. 6.
In-house lawyers and compliance managers in regulated sectors who need a defensible, signed classification on file before the 2027 deadline.
Technology leaders who need to present a board-ready classification position and understand the downstream obligation cascade before it becomes urgent.
Product managers and AI owners in regulated industries who need to scope compliance obligations before building, procuring, or deploying a new system.
Risk management (Art. 9), data governance (Art. 10), technical documentation (Art. 11), logging (Art. 12), transparency (Art. 13), human oversight (Art. 14), and accuracy (Art. 15) — all of these are triggered by a high-risk classification. Until the classification is done, you cannot scope your obligations. The Fundamental Rights Impact Assessment under Art. 27 is also conditional on this determination.
Purchase includes the complete 8-chapter guide and the companion Excel register. Instant download via ClickBank's secure checkout. No subscription. No renewal.
Complete guided classification instrument for EU AI Act Article 6 and Annex III.
Governance Academy members receive this guide as part of their membership toolkit. Log in to your member dashboard to download it — no additional purchase required.
The 2 December 2027 deadline approaches faster when you factor in the time it takes to inventory your systems, complete the test, and get sign-off from stakeholders. The guide is designed to be completed in one sitting.
GET THE GUIDE — £47 →