Where does your AI system sit under the EU AI Act?
A first-pass classification, the obligations that follow from it, and the evidence to ask your vendor for — kept as an AI system record you can file.
16 questions, fewer where some do not apply, on your role, the EU connection, the use case, who is affected, the oversight in place and the practices the Act forbids outright. You leave with a risk tier and a stated confidence, your likely role under the Act, the duties that follow, the vendor evidence you are missing, and the triggers that should send you back for a reassessment.
The tier is set by versioned rules run against the consolidated text of the Regulation, pinned at a named CELEX and current to 27 July 2026. No language model decides your classification. Every obligation the result shows is anchored to the provision it rests on, and each of those 28 provisions is readable here in the pinned text.
The Article 50 transparency duties and the penalty regime have applied since 2 August 2026. The next dated step is 2 December 2026: new prohibitions, and machine-readable marking. The result tells you which line your system sits on.
- Step 1
Describe it in your own words
What the tool is, who it affects and what happens with its output. Or skip straight to the questions.
- Step 2
Confirm what we read
The description becomes draft answers. You accept or reject each one; nothing enters the engine unconfirmed.
- Step 3
Answer what is left
The remaining questions are asked, not assumed — the prohibited-practice screen is always among them.
- Step 4
Keep the record
Tier, role, duties, vendor questions, review triggers and the rules that fired, with the Article behind each.
This is a first-pass triage against the text of the Act, not legal advice. A high-risk or prohibited result is a reason to take advice, not a finding against you.
What a result looks like
Three systems, run through the same rulesA third-party tool ranks applicants; a recruiter reviews the shortlist but usually goes with it.
Customers in the EU talk to an assistant built on a vendor model; it answers questions and hands off to staff.
Staff use it for first drafts and research; every output is rewritten before it leaves the building.
You are interacting with an AI system. It reads your description and proposes answers to the questionnaire. It does not decide your risk tier — that stays with the rule-based engine, and you confirm every answer first.
AI Act Compliance Toolkit
From £79Your Checker result is a first pass over one system. The Toolkit is what you work the whole estate through: assess gaps, catalogue every AI system, classify risk, score supplier dependencies and build an evidence-backed action plan.
Advisory Briefing
£450Which actions matter most, and what do you need to establish next? Work through your result with an independent adviser, grounded in how your business uses the system.
The output of a tool or product, or one of your AI systems. A one-hour discussion and a written follow-up with priorities, evidence gaps and next actions.