SAMPLE REPORT — fictional company, fictional system, fictional findings. A real report is written from your own questionnaire answers.

EU AI Act Classification Report

Sample Company — example · System: “SampleHire Screen” · Prepared from a sample questionnaire

Report cover SAMPLE

Prepared forSample Company — example (samplecompany.example) — a fictional company invented for this sample. Any resemblance to a real business is a coincidence.
System classifiedSampleHire Screen — a fictional AI tool that reads job applications, scores candidates against a job description and drafts a shortlist for a human hiring manager.
BasisThe customer's answers to the 30-question classification questionnaire (answers invented for this sample).
Reference lawRegulation (EU) 2024/1689 (the EU AI Act), public text, read as of the sample date.
DeliveredBy email, as a written report — exactly like this page, minus the SAMPLE marks.

This sample shows the exact structure and depth of a real Classification Report: verdict, reasoning section by section, obligations mapped to the category, deadlines, penalty exposure and a step-by-step action plan. The company, the system and every finding below are invented.

1 · Verdict at a glance SAMPLE

HIGH-RISK

Annex III, point 4(a) — AI used for recruitment and selection of natural persons. The full high-risk obligation set applies to this sample system.

2 · The system we classified SAMPLE

From the sample questionnaire answers:

3 · Reasoning, section by section SAMPLE

Step 1 — Is it banned? (Article 5)

No. The sample system does not use subliminal or manipulative techniques, does not exploit vulnerabilities, is not a social-scoring system, and does not perform real-time remote biometric identification in public spaces. Nothing in the answers matches the prohibited list — so the analysis moves to the risk tiers. Had the tool scored candidates' trustworthiness across unrelated contexts, this step could have ended differently — that is why the questionnaire asks about it directly.

Step 2 — Is it high-risk? (Article 6 and Annex III)

Yes. Annex III, point 4(a) lists AI systems intended to be used for the recruitment or selection of natural persons — including analysing and filtering job applications and evaluating candidates. SampleHire Screen does exactly that. Article 6(2) makes Annex III systems high-risk. The Article 6(3) filter (a system that does not significantly affect the outcome) does not rescue this sample: its scores decide who reaches the shortlist, which is a material influence on who gets interviewed — and it performs profiling of natural persons, which keeps it high-risk regardless.

Step 3 — Do transparency duties apply? (Article 50)

Yes, in parallel. Candidates interact with an AI system and are entitled to know. The sample product tells candidates only in its terms of service — the report flags this as a gap: disclosure must be clear and at the point of interaction, not buried in terms.

Step 4 — The GPAI track

The underlying model is a general-purpose AI model from a third-party provider. GPAI duties (technical documentation, information to downstream providers, and for systemic-risk models, evaluations) sit with that model's provider and have applied since August 2025. They do not replace Sample Company's duties for its own high-risk system.

4 · Obligations that follow SAMPLE

For a high-risk system, the provider (Sample Company) must have in place, before the duties apply:

Deployers (the employers using it) carry their own duties under Article 26 — use per instructions, competent human oversight, relevant input data, log retention — and, for Annex III systems used on workers or candidates, must inform the people affected (Article 26(7)). A fundamental-rights impact assessment (Article 27) applies to specific deployer types; a typical private employer using this sample tool is outside that list, but public-sector deployers may not be.

5 · Deadlines SAMPLE

From 2 February 2025Prohibited practices banned; AI-literacy duty applies. (In force.)
From 2 August 2025General-purpose AI model duties apply. (In force — relevant to the sample's model provider.)
From 2 August 2026Most remaining obligations apply, including the Annex III high-risk duties that govern this sample system. The EU has been discussing postponing some high-risk deadlines, so exact dates may shift — plan against August 2026, not against a postponement.
From 2 August 2027High-risk duties for AI that is a safety component of regulated products (Annex I) — not this sample system.

6 · Penalty exposure SAMPLE

For a small company the fixed amounts are theoretical ceilings — but enterprise customers, insurers and partners ask for the classification and the plan long before any regulator does. That is the practical exposure.

7 · Step-by-step action plan SAMPLE

  1. Fix the logging gap first (Article 12): keep scores and reasons for all candidates, not only the shortlisted — you cannot audit what you did not record.
  2. Move candidate disclosure up front (Article 50): a plain notice where candidates apply — "an AI system helps rank applications; a person makes the final decision."
  3. Write the technical documentation (Annex IV) from the questionnaire answers and your system design — purpose, data, accuracy, known limits.
  4. Run a bias and data-governance review (Article 10): test scores across protected groups on historical applications before the next release.
  5. Formalise human oversight (Article 14): name who can override, train them, and log overrides.
  6. Stand up the risk-management and quality files (Articles 9 and 17) — one living document each, reviewed on every significant change.
  7. Prepare the conformity assessment, EU declaration and database registration (Articles 43, 48, 49) ahead of the August 2026 application date.
  8. Ship deployer instructions (Article 13) with the product: how to use it lawfully, what to tell candidates, how long to keep logs.

Honest limits SAMPLE

This report is an educational classification based on the public text of the EU AI Act and on the answers given in the questionnaire. It is not legal advice, not a conformity assessment and not certification, and it does not replace a lawyer. If an answer in the questionnaire is incomplete or a system's real use differs from its description, the classification can change — the Act classifies what a system does in deployment, not what its vendor calls it. SkillQuesters is independent and is not affiliated with the European Commission.

Your system, classified the same way

Answer the same 30 questions about your real system and receive a report with this exact structure — your verdict, your obligations, your plan — by email within 3 business days of your completed questionnaire.

Get My Report — $49

Back to the EU AI Act Classifier · Or send the questionnaire free first

SAMPLE — fictional company, fictional system, fictional findings. Your report is generated from your real questionnaire answers and delivered by email.