Free tool · no sign-up, nothing stored, nothing sent anywhere
Which Article 50 duties
might already apply to you?
The EU AI Act's transparency obligations have applied since 2 August 2026. Most coverage has been about labelling AI-generated content, which is the narrow one. The rule that catches far more organisations is simpler: if a system talks to a person, that person has to know it is a machine.
Six questions. Everything runs in your browser. This flags what is worth looking at. It does not decide anything, and it cannot tell you that you are compliant.
-
Do any of the people who use, or are affected by, your AI system sit in the EU?
The Act follows the people, not your registration. A UK company with EU users is in scope; an EU company whose system only ever touches people outside the EU may not be.
-
if you build it
Do you build or supply an AI system that interacts directly with people?
A support chatbot, a voice agent, an assistant embedded in your product.
Would it be obvious to an ordinary observant user that they are talking to a machine, without being told?
Read this narrowly. A widget labelled only “Chat” is not obviously AI, and neither is a named persona.
-
if you build it
Do you build or supply an AI system that generates audio, images, video or text?
Anything producing synthetic content, including a general-purpose model you offer to others.
Does it only assist with standard editing, without substantially altering the input or its meaning?
Spellcheck and colour correction sit outside. Rewriting, summarising or generating does not.
-
if you use it
Do you publish or use AI-generated or manipulated images, audio or video that could be taken for real?
Synthetic presenters, voice cloning, edited footage of real people or events.
-
if you use it
Do you publish AI-generated text in order to inform the public on matters of public interest?
News, analysis, public health or safety information. Ordinary marketing copy is not this.
Has it been through human review, with a named person or organisation holding editorial responsibility?
This exemption is explicit in the text and it is how most publishers will land outside the obligation.
-
if you use it
Do you use a system that infers emotions, or categorises people using biometric data?
Sentiment from voice or face, biometric categorisation in hiring, retail or security.
What this is, and what it is not
This is not legal advice and I am not a lawyer. It reads the text of Article 50 and points at the paragraphs your answers suggest are worth looking at. It does not cover the rest of the Act, and it cannot tell you whether your particular system is a deep fake, or whether your content informs the public on a matter of public interest. Those are judgement calls, and they are the ones worth taking to somebody qualified.
It cannot tell you that you are compliant, and a result with nothing flagged is not a clean bill of health. It is deliberately biased towards raising things: being pointed at an obligation that turns out not to apply costs you a conversation, and being told you are fine when you are not costs considerably more.
Source: Regulation (EU) 2024/1689, Article 50. Article 50 was not deferred by the 2026 omnibus package that moved the Annex III high-risk deadlines. Checked 4 August 2026. Read the Regulation.