Blog

The EU AI Act gets real: what does Article 50 mean in practice?

The EU AI Act has been talked about for a long time. Now it finally means something. And it means something for all of us who build or use AI-based digital services. On 2 August 2026, the Act's transparency obligations in Article 50 started to apply. In practice, they affect every organisation that uses AI in its website, online service or content.

Author: Aapo Mäki, CEO / Senior Service Designer, partnerPublished Updated 3 min read

This is no longer only about the big AI providers, such as the giants behind Claude and ChatGPT. This is about you, if you use an AI chatbot in customer service. About you, if you produce content with AI. About you, if your website has any AI-based feature or gadget at all. Which means us too.

What is it about?

Article 50 (opens in a new tab) of the EU AI Act sets out four situations that trigger a transparency obligation:

First: when AI interacts directly with a person, such as a chatbot or a virtual assistant. The user must be told that they are talking to an AI. Right at the start, clearly. Not hidden in the terms of use.

If your website has images or videos generated by AI, they must be marked as AI-generated.

Second: AI tools that generate synthetic content, meaning text, images, audio or video. Such content needs at least a machine-readable marking that says it was generated by AI. The EU is currently developing a standardised “AI label” for this purpose. At this point, the marking is the responsibility of the company that provides the AI tool.

Third: when AI is used for things like emotion recognition or biometric categorisation. The people concerned must be informed. It will be interesting to see how this works in practice…

Fourth: deepfakes and AI-generated text that aims to inform the public on matters of public interest. Perhaps like this piece, which was written with help from Claude. In that case, the AI origin of the content should be disclosed, unless a human has genuinely reviewed it editorially.

What does this mean for an ordinary organisation?

  1. If your organisation has a chatbot on its website, check whether it says that it is an AI. If it doesn't, fix it by August.
  2. If you produce newsletters, blog articles or other content with AI assistance and publish them to the general public, the obligation applies to you. Taking editorial responsibility does exempt you, but it has to be genuine, not a formality.
  3. If your website has images or videos generated entirely by AI, they must be marked as AI-generated.

The key point is that Article 50 is not limited to so-called high-risk AI systems*. It covers every practice and tool that falls into the four situations above. So many organisations that have no high-risk AI system at all will still find themselves within the scope of this obligation. We are one of them.

*High-risk AI systems in brief: in the AI Act, “high-risk” means applications used for certain purposes, which are regulated more strictly than the four themes in this article. The list includes, for example:

  • HR and recruitment: AI that scores or screens job applicants
  • Lending: AI that assesses creditworthiness
  • Healthcare: diagnostic tools
  • Education: systems that assess students
  • Critical infrastructure: for example energy and transport
  • Justice: systems that support decision-making
  • Police and border control: risk assessment and identification

AI? Fascinating. I love it.

Best regards, Aapo

Read more

All articles
  • Blog

    I built a website for five grand, but it would cost you 30 grand

    We are one of Finland's leading WordPress agencies. Yet our new website does not contain a single line of WordPress. Most of the site was built by me, the CEO, and I can't code. In this article I explain what that proves and what it doesn't.

    19 Nov 2026 · 8 min read

  • Blog

    WordPress security vulnerabilities: what is actually going on?

    July and August 2026 were a historically heavy stretch for the WordPress ecosystem. Within a couple of months we saw a critical vulnerability in WordPress core itself, more than a hundred plugin security fixes, an unpatched login bypass in a popular SSO plugin and a case where malicious code reached thousands of sites straight through the official plugin directory. The National Cyber Security Centre Finland at Traficom also issued its own warning. In this article we explain in plain language what happened, why and when.

    17 Aug 2026 · 13 min read

  • Blog

    How does a website move business forward?

    We say we build websites that move our clients' business forward. What does that actually mean? In practice, we first agree on the goals together with the client, such as growing the number of leads. Only then do we design and build what serves those goals. Finally, we follow the numbers and keep improving the site based on the results.

    13 Feb 2026 · 2 min read

Get a price estimate or tell us what you need

The estimate takes a few minutes and does not ask for your contact details. You can also describe your need. We will get back to you by the next working day at the latest.

Get a priceTell us what you need