Transparency · Article 50
The AI Act now expects you to tell people when they are dealing with AI, and to keep proof you did. Normis works out whether Article 50 applies to you, hands you the wording to use, and keeps a dated record for when someone asks.
Who it is for
Most small companies are not sure Article 50 is even their problem. The honest answer: it depends on whether your AI faces your customers, not on how big you are. Article 50 is about outward-facing AI.
You likely have an Article 50 duty if any of these are true:
You probably do not have an Article 50 duty if your only AI use is internal - drafting with ChatGPT, code assistance, an AI notetaker in meetings. That use can still touch other parts of the Act, such as the Article 4 AI-literacy expectation, and the free classifier covers that too. Still weighing it up? See what actually applies to you.
Article 50 wants the disclosure clear and given at the first interaction, not buried in your terms and conditions.
Your transparency register is the list of where AI meets your customers, and what you did about it.
Your transparency entries fold straight into the Normis evidence pack, the single sealed document that holds your records together.
How it works
Run the free classifier; it flags which of your tools trigger Article 50.
Paste the right disclosure template into your chatbot, content or media.
Log each outward-facing AI in the transparency register and attach a screenshot.
Your disclosures come out inside the sealed evidence pack, ready for an auditor or customer.
If any of your AI is customer-facing - a chatbot, AI-generated marketing content, synthetic media, or emotion and biometric tools - then yes, Article 50 likely applies, whatever your company size. If your AI use is purely internal, Article 50 generally does not apply, though other parts of the Act might. The free classifier gives you a clear answer per tool.
The transparency obligations apply from 2 August 2026, alongside national enforcement powers. It is one of the first parts of the AI Act that is actually live for ordinary businesses.
Article 50 is about telling other people they are dealing with AI, so internal-only use usually falls outside it. That use can still touch the Article 4 AI-literacy expectation; the classifier covers that too, so you see the whole picture, not just Article 50.
Any AI system meant to interact directly with people - a support chatbot, a website assistant, an automated phone agent. People must be told they are dealing with AI at the first interaction, unless it is already obvious to a reasonable person.
No. The Regulation expects the disclosure to be clear and noticeable at the point of interaction. A notice buried in terms, a manual, or a layered menu does not meet the standard. That is why Normis gives you a short notice to show up front.
If the content is artificially generated or manipulated - including synthetic or deepfake media - it generally needs to be disclosed as such, and marked in a machine-readable way. Normis gives you a label template and a watermarking checklist so you do not miss the technical part.
From 2 August 2026, national authorities have enforcement powers, and Ireland's AI Office is operational. No fines have been issued across the EU so far, but the duty is live, and increasingly enterprise customers ask for proof during procurement before any regulator does.
No. Normis gives you plain-English guidance, disclosure templates and a record-keeping system, with the article reference kept next to every answer. It is not a substitute for legal advice on a complex or high-stakes case.
The classifier is free, no signup, no card. It tells you whether Article 50 applies to you, and what to do about it. Just a clear answer and a calmer week.