Chatbots and generated content: the transparency duty
Since 2 August 2026, Article 50 of the AI Act has set transparency rules. Someone chatting with a chatbot must know that they are talking to an AI, unless it is obvious. An AI-generated image, sound or video that could pass for real must be labelled. The information must be clear and given at the first contact at the latest.
What Article 50 says
Article 50 of Regulation (EU) 2024/1689, the AI Act, sets transparency obligations. They have applied since 2 August 2026. Their aim: people should know when they are dealing with an AI.
The text contains several rules:
- Interacting with an AI (paragraph 1). People must be “informed that they are interacting with an AI system”, unless this is obvious.
- Technical marking (paragraph 2). Tools that generate text, images, sound or video must mark what they produce. This marking must be detectable by a machine. This rule is aimed at the companies that make the tools. Tools already on the market before 2 August 2026 have until 2 December 2026 to apply it.
- Emotion recognition (paragraph 3). People exposed to this type of system must be told about it.
- Deepfakes (paragraph 4). A deepfake is an image, sound or video that resembles real people, places or events. It could pass for authentic. Whoever publishes it must state that it was generated or manipulated by AI.
- Texts on matters of public interest (paragraph 4). An AI-generated text published to inform the public on matters of public interest must be labelled. The exception: it has been reviewed by a person who takes editorial responsibility for it. A simple spelling check does not count as a review.
The information must be given “in a clear and distinguishable manner at the latest at the time of the first interaction or exposure”. It must also be accessible.
Content generated before 2 August 2026 does not have to be labelled after the fact. The European Commission still encourages doing so where possible.
Who is concerned
The regulation distinguishes two roles. The provider develops the AI system and places it on the market under its own name. The deployer uses it at work. Most SMEs are deployers.
The paragraph 1 obligation falls on the provider: it must design the chatbot so that it informs the user. If you have a chatbot developed and put into service under your own name, you may be considered the provider. Either way, it is your website and your customers: check that the information really appears.
The deepfake obligation falls directly on the deployer. If your business publishes such content, it is up to you to label it. The invisible marking added by the tool is not enough: the label must be visible or audible.
Three examples for an SME
A chatbot on your website
Your website offers an assistant that answers questions about opening hours or delivery times. Display a message as soon as it opens, for example: “You are chatting with an automated assistant based on AI. An adviser can take over.” If the chatbot collects personal data, the GDPR (the EU data protection regulation) also applies: inform the people concerned.
Generated images in your adverts
You create a campaign visual with an image generation tool. If it shows a realistic person, place or event that could pass for real, add “AI-generated image”. For an obviously artistic or fictional creation, the text allows a more discreet notice that does not spoil the work. If in doubt, label it.
A synthetic voice
Your phone line uses a generated voice to answer calls. Announce at the start of the call that it is an automated assistant. If the voice imitates a real person, it is an audio deepfake: label it, and get that person’s consent beforehand.
The essentials in one table
| Situation | Who must act | What you do |
|---|---|---|
| Chatbot on the website | The provider, and you in practice | Visible message from the first interaction |
| Realistic generated image | You, as the deployer | “AI-generated image” label |
| Synthetic voice on the phone | The provider, and you in practice | Announcement at the start of the call |
| Blog post written with AI and reviewed | Usually not concerned | Review and editorial responsibility |
Find out more
These rules come on top of the GDPR: they do not replace it. For the details, the European Commission published guidelines on 20 July 2026. Train the people who publish content or manage the website with the Learn path. Review your situation with the self-assessment.
Frequently asked questions
Do I have to disclose an email written with the help of an AI assistant?
Article 50 does not cover this case: the email is sent by a person, who remains responsible for it. Always reread it before sending, because the assistant can make mistakes.
Is the technical marking of generated content up to me?
No, paragraph 2 is aimed at the providers of generation tools. For tools placed on the market before 2 August 2026, providers have until 2 December 2026 to comply. This invisible marking does not replace the visible label you must add to a deepfake.
How should I word the notice on a chatbot?
A short sentence is enough, visible as soon as the chatbot opens: “You are chatting with an automated assistant based on AI.” Also say how to reach a person, which reassures your customers.
Sources checked
- AI Act Service Desk (European Commission): Article 50, French version(opens in a new tab)
- AI Act Service Desk (European Commission): Article 50, Transparency obligations (consolidated version)(opens in a new tab)
- AI Act Service Desk (European Commission): Article 111, Systems already placed on the market(opens in a new tab)
- AI Act Service Desk (European Commission): Article 3, Definitions(opens in a new tab)
- AI Act Service Desk (European Commission): Article 113, Entry into force and application(opens in a new tab)
- CNIL (French data protection authority): questions and answers on using a generative AI system, in French(opens in a new tab)
- EUR-Lex: Regulation (EU) 2026/1744 (Digital Omnibus on AI), which amends the AI Act(opens in a new tab)
- European Commission: guidelines on the transparency obligations of Article 50(opens in a new tab)
- European Commission: questions and answers on the transparency obligations of Article 50(opens in a new tab)
This guide gives information, not legal advice. Check the official text in force for your situation.