Article 50 in four lines.
| Paragraph | Who | Obligation, simplified |
|---|---|---|
| 50(1) | Providers of AI that talks to people | Design it so people are told they are interacting with AI, unless obvious. |
| 50(2) | Providers of generative AI | Mark AI-generated audio, image, video or text output in a machine-readable way. As reported, systems already on the market before 2 August 2026 have until 2 December 2026. |
| 50(3) | Deployers of emotion recognition or biometric categorisation | Inform the people exposed to it. |
| 50(4) | Deployers of deepfakes, or of AI text published to inform the public | Disclose that the content is AI-generated or manipulated, with exceptions. |
Simplified reading of Article 50 for orientation; see the official text for the exact conditions and exceptions.
How we implement it per channel.
The first reply names the assistant: for example, an AI assistant of your business, and how to reach a colleague.
A label on the chat window plus the same opening line; the hand-off button is always visible.
The voice agent says it is an AI assistant before anything else, and transfers on request.
Drafts approved and sent by staff are your staff's messages; fully automatic replies say they are automated.
Disclosure and hand-off requests are logged, so you can show how it works.
The same standard in the Gulf.
UAE, Saudi and Qatari law do not have an Article 50 equivalent written in the same way. We apply the same disclosure anyway: customers in Dubai notice when a reply comes at 2 a.m., and saying it is an assistant avoids the feeling of being misled.
In the Netherlands, the AP and the ACM said in October 2025 that organisations must make clear when someone is dealing with a chatbot, must not give wrong or evasive information through it, and must always offer the option of speaking to an employee.
Where this comes from.
- Regulation (EU) 2024/1689 (Artificial Intelligence Act), EUR-Lex
- AI Act Explorer: Article 50, transparency obligations (unofficial reading aid)
- ActuIA: AI Act obligations in force and delays under omnibus 2026/1744 (secondary summary)
- NJB: AP and ACM say customer contact may not run only through an AI chatbot (2 October 2025)
Not legal advice.
This guide is general information, checked against the sources listed, and not legal advice. Rules change; check the current text or ask a lawyer before you rely on it. Checked November 2026.
Common questions.
When is it 'obvious' that it is AI?
The Act gives the exception, not a list. A clearly labelled voice assistant may be obvious; a WhatsApp message that reads like a colleague is not. We do not rely on the exception.
Does this apply to a business in Dubai?
It applies when the system is used with people in the EU. A Dubai business serving only UAE customers is outside it, but the practice is still sensible.
Is there a fine?
The AI Act sets penalty tiers, and transparency obligations fall under them. This is general information, not legal advice.
Does Article 50 make our chatbot high-risk?
No. Transparency obligations are separate from the high-risk rules. A customer-service agent is usually not high-risk.
Want this in your business?
Tell us how you work today. We'll tell you honestly what AI can take off your team's plate, what it costs, and whether we are the right fit.