In some cases, yes, since August 2, 2026 (EU AI Act, Article 50): a chatbot must say it is an AI, and realistic AI-generated or AI-altered content (deepfakes), as well as AI-generated text published to inform the public on matters of public interest, must be disclosed. An email or a sales listing reviewed by a human does not require a systematic label.
What Article 50 says.
The EU AI Act includes transparency obligations that apply since August 2, 2026. The "omnibus" regulation adopted in July 2026, which postponed other deadlines, did not postpone these. They cover four situations:
- Interacting with an AI: a person who interacts with an AI system (chatbot, voice agent) must be informed, unless it is obvious from the context.
- Marking synthetic content: providers of tools that generate text, images, audio or video must mark it in a machine-readable way. A transition period runs until December 2, 2026 for systems already on the market. This obligation falls on the vendors (Anthropic, OpenAI…), not on you.
- Deepfakes: whoever publishes a realistic image, audio or video generated or altered by AI, which could pass as authentic, must disclose it.
- Public-interest text: AI-generated text published to inform the public on matters of public interest must be disclosed, unless it has undergone human review or editorial control under someone's responsibility.
In short: the AI Act does not ask you to label everything AI helped produce. It asks you not to mislead people about whether they are talking to a machine or looking at fabricated content.
Case by case, in a small business.
| Situation | Disclosure required? | In practice |
|---|---|---|
| Chatbot on your website | Yes | It introduces itself as an AI in the first message |
| Voice agent on the phone | Yes | Announcement at the start of the call, transfer to a human available |
| Realistic generated visual (virtual staging, product scene) | Yes | A visible label on or next to the image |
| Email or quote drafted with AI, reviewed and sent by an employee | No, not under the AI Act | The employee remains the author and is responsible for the message |
| Product page, post or sales newsletter, reviewed | Not systematically | Human review and editorial responsibility |
| Informational article published without review | Yes if the topic is of public interest | Better to review it, it is also a quality issue |
Real estate is a good example: since August 2, 2026, realistic virtual staging must be disclosed as such, and editing must never hide a defect in the property.
How to display it, simply.
- Chatbot: a welcome line ("I'm [company]'s virtual assistant, an AI. I can make mistakes; a person can take over.") and a button to reach a human.
- Voice agent: the same announcement, spoken at the start of the call.
- Visuals: a short label ("AI-generated image", "Virtual staging") where the image is seen, not hidden in the legal notice.
- Website: a paragraph in your privacy policy if AI processes visitors' personal data.
And GDPR?
The AI Act adds to the GDPR; it does not replace it. As soon as a chatbot or an agent processes personal data (name, email, the content of a request), the usual rules apply: informing people, data minimisation, retention periods, identified processors. The French data protection authority (CNIL) notes that an organisation connecting an AI model to a knowledge base containing personal data becomes responsible for that processing.
What I recommend.
Go a little beyond the legal minimum, because it pays off: a customer who discovers on their own that they were talking to a machine loses trust, whereas an assistant that introduces itself clearly and hands over at the right moment is generally well accepted. On the agents I build, the "I'm an AI" announcement and the hand-off to a human are part of the specification from the pilot onwards.
For your content, keep one simple internal rule: everything that goes out in the company's name is reviewed by a person. You respect the spirit of Article 50 and avoid the factual errors AI can produce.
Related questions.
Is there a mandatory wording?
No single wording is required. What matters is that the information is clear, visible, and given at the latest at the first interaction or exposure to the content. "I'm the virtual assistant of [company], an AI" works perfectly for a chatbot.
Do AI-retouched product photos need a label?
Light editing (brightness, cropping, background removal) does not create misleading content. But a realistic image showing something that does not exist, such as virtual staging or a product in a generated setting, follows the deepfake logic and should be disclosed. When in doubt, label it: a label costs little, lost trust costs a lot more.
Sources
- European Commission : Regulatory framework on AI, AI Act timeline (page updated August 3, 2026)consulted October 13, 2026
- EUR-Lex : Regulation (EU) 2026/1744 of July 8, 2026 (AI omnibus), in force since July 27, 2026consulted October 13, 2026
- Houthoff : AI Act transparency obligations and the AI omnibus regulationconsulted October 13, 2026
- iubenda : AI Act transparency obligations (July 10, 2026)consulted October 13, 2026
Information last checked on October 13, 2026. General information, not legal advice: for a specific case, ask your data protection officer or a lawyer.