The EU AI Act went live on 2 August. Does it apply to you?
Seven questions, two minutes. You will get the short list of what you actually need to do, including the exact disclosure lines you can copy and paste.
This is a practical checklist, not legal advice. If you operate in a regulated sector or anything below raises a red flag, have a lawyer confirm.
1. Do you have customers or users in the EU? (Including EU visitors using your website tools or reading your content.)
2. Do you run a chatbot, AI assistant, or AI voice agent that customers talk to?
3. Do you publish AI-generated images, audio or video? (Realistic ones a person could mistake for a real photo, voice or recording.)
4. Do you publish AI-written articles or news-style content without a human reviewing and taking editorial responsibility?
5. Do you use AI for emotion recognition or to categorise people by biometrics (faces, voices)?
6. Does your AI do any of: credit or creditworthiness scoring, CV or candidate screening, education scoring, or decide access to essential services?
7. Do you ever publish deepfakes, meaning realistic AI content showing real people or events?
Prefer to check with your own AI?
Paste this into ChatGPT or Claude and answer its questions. It uses the same verified rules as this page.
You are an EU AI Act quick-checker for a UK/EU SME. Today the Article 50 transparency obligations are in force (live since 2 August 2026). Interview me one question at a time: (1) EU customers or users? (2) customer-facing chatbot or voice AI? (3) published AI-generated images/audio/video that could be mistaken for real? (4) AI-written public-facing articles without human editorial review? (5) emotion recognition or biometric categorisation? (6) credit scoring, CV screening, education scoring or essential-services access decisions? (7) published deepfakes of real people or events? Then tell me, in plain English: which Article 50 duties apply to me and the exact one-line disclosures to add; whether anything falls under the high-risk Annex III list (deadline 2 December 2027) and needs a lawyer; and what is noise I can ignore. Rules you must respect: fines for transparency breaches cap at 15m EUR or 3% of turnover (the scary 35m EUR / 7% tier is for prohibited practices only, and for SMEs the lower figure applies); machine-readable watermarking is the AI tool provider's job, not mine, with a grace period to 2 December 2026 for systems already on the market; a stylised graphic is not a deepfake. Keep it under 300 words, end with "this is not legal advice".
Built by OptiBee, the AI build studio for UK SMEs. We build the disclosures in from day one.
Want your systems checked properly against this, or an AI build that is right by default? Talk to us.