Does our chatbot have to say it is an AI? Article 50 explained
Short answer
Yes. From 2 August 2026, a chatbot must be designed so that people can see they are talking to an AI, unless it is obvious (Article 50(1) – a duty for the provider of the chatbot, which is often you if you built it). Generative AI must mark its output in a machine-readable way (Article 50(2)). If you use deepfakes – images, audio or video that resemble real people or events – you must disclose that the content is artificial (Article 50(4)). The same applies to AI-generated text published to inform the public, unless a person has edited it and takes responsibility for it.
Four situations in Article 50
| Situation | Who has the duty? | What must happen? |
|---|---|---|
| People interact with an AI | The provider of the AI system | It must be clear that it is an AI |
| AI generates text, images, audio or video | The provider of the generative AI | Machine-readable marking |
| Emotion recognition or biometric categorisation | The deployer of the system | The people affected must be informed |
| Deepfakes and AI text for the public | The deployer of the system | It must be disclosed that the content is artificial |
Did you build the chatbot yourselves?
If you have built a chatbot on your website using a language model from, say, OpenAI or Anthropic, you are typically the provider of the chatbot – and the duty in Article 50(1) is yours. If you use a ready-made chatbot from a vendor, it is the vendor's duty to make it possible, but in your interest to check that it works.
GoalCube has ready-made templates for a chatbot notice and for labelling AI content.
What to do
- Find your chatbots.On your website, in your app, in customer service and on social media.
- Put a clear notice at the top.For example: “You are chatting with an AI assistant. Type ‘agent’ to reach a person.”
- Label deepfakes.Images, video and audio that resemble real people must carry a visible label.
- Check your provider.Generative AI must mark content in a machine-readable way. That is the provider's duty, but ask.
What the law says
- Article 50 AI Act – transparency obligations
- Applies from 2 August 2026. Systems placed on the market before that date have until 2 December 2026 for machine-readable marking (Article 50(2)).
- Article 99(4) AI Act
- Breaches of Article 50 can lead to fines of up to €15 million or 3%.
Questions and answers
What counts as “obvious”?
When a reasonably well-informed person can tell from the context. If in doubt, say it.
Do AI-written newsletters need labelling?
Only text published to inform the public on matters of public interest, and only if it hasn't been edited by a person who takes responsibility for it. An ordinary newsletter about your products is rarely covered.
Do AI images used in marketing need labelling?
If they are deepfakes – that is, they resemble real people, places or events and could be taken as genuine – yes. Artistic or clearly fictional images have lighter requirements.
Sources
Written by The GoalCube editorial team. Last updated 11 October 2026. GoalCube helps you follow the EU AI Act and GDPR. It is not legal advice.
Be among the first.
We expect to launch in November 2026. Join the waitlist and we’ll let you know when you can get started.