NewRAGSuite Engine: the knowledge engine for your data, run in Germany.About RAGSuite Engine

ProductRAG enginePlatform overviewAI governanceAI chatbotBuild AI assistantsAI searchAI knowledge baseConnectors & integrationsMCP serverTake the product tour →
Engine & self-hostedRAGSuite EngineSelf-hosted / on-premiseOpen source (Community)RAGSuite Engine or self-hosted? →
SolutionsSMEsHealthcarePublic sectorBanking & insuranceChatGPT alternativeAgencies & partnersAll solutions →
Case studies
Pricing
ResourcesGuidesBlogFree toolsAI data protectionEU AI ActTrust Center
Open source on GitHub
Start free trialBook a demo
Free tool · no sign-up · runs in your browser

AI labelling: the right notice for chatbots, text and images.

Three questions, one ready notice: the generator writes AI disclosures for chatbots, AI text, AI images, marketing and your website, in German and English. It also tells you whether Article 50 of the EU AI Act requires a label and where the notice should go.

AI labellingQuestion 1 of 3
UseWhat do you want to label?
Nothing is stored
Your notice

Your AI notice, ready to copy.

Answer the three questions above. Your notices in German and English then appear here, with the Article 50 check and where the notice should go.

In plain words

What does AI labelling mean, and when is it required?

AI labelling makes it visible that content or a conversation partner comes from an AI. Since 2 August 2026 it has been required in certain cases: Article 50 of the EU AI Act sets out four triggers, regardless of risk class and for open-source systems too.

The table shows who has to label what. Outside these cases a notice is voluntary but often sensible, because it builds trust.

CaseDuty under Art. 50Who
AI talks to people (chatbot, voice assistant, AI agent)Notice at first contact at the latest, unless obvious (para. 1)Provider
AI generates text, images, audio or videoMachine-readable marking (para. 2)Provider
Emotion recognition, biometric categorisationInform the people concerned (para. 3)Deployer
DeepfakesDisclose that it is artificially generated (para. 4)Deployer
AI text on public-interest topicsDisclose, unless there is editorial review and responsibility (para. 4)Deployer

Source: Art. 50 Regulation (EU) 2024/1689; European Commission guidelines on transparency obligations of 20 July 2026. For generative systems placed on the market before 2 August 2026, the marking duty under para. 2 applies from 2 December 2026.

The AI labelling duty under Art. 50 →

What is not enough

Three notices that the guidelines say are not enough.

Only in the terms

A notice nobody sees during the conversation is not a notice.

Only a watermark

Machine-readable marking does not replace telling the person.

Only “assistant”

The word alone does not tell anyone that it is an AI.

How to proceed

Four steps to clean AI labelling.

01

List your uses

Where do you use AI: chat, text, images, marketing? A list is enough.

02

Check the duty

Check for each use with this generator whether Art. 50 applies.

03

Add the notice

Place the text where people meet the AI, not only in the terms.

04

Record the rules

Set out labelling in your AI policy and train the team.

Create an AI policy →AI notice in the chat widget →

FAQ

Questions about AI labelling.

What does AI labelling mean?

AI labelling means making it visible that content or a conversation partner comes from an AI, for example with a notice in the chat window, an image caption or a line below a text.

Is there a legal duty to label AI?

Yes, since 2 August 2026 under Article 50 of the EU AI Act, but only in certain cases: chatbots and other AI that talks to people, deepfakes, AI text on matters of public interest without editorial review, and emotion recognition. Generative AI systems must also mark their output in a machine-readable way.

Is a note in the terms enough?

No. According to the European Commission’s guidelines of 20 July 2026, a note in the terms, a watermark alone or the vague label “assistant” is not enough. The notice must be visible where people meet the AI.

Do I have to label AI images?

For deepfakes, yes: if an image shows real people, places or events in a deceptively real way, you must disclose that it is artificially generated. For other AI images there is no explicit duty for deployers, but a notice is good practice and platforms often require it.

What happens without a label?

Breaches of Article 50 can be fined up to €15 million or 3% of worldwide annual turnover. In Germany, market surveillance lies with the Bundesnetzagentur.

Are the templates legally watertight?

The templates are a starting point as of 30 October 2026, not legal advice. Whether and how you must label depends on your use and your role; if in doubt, check with your legal adviser.

Stefan Reinhardt
Stefan Reinhardt
Public Sector & Compliance

A chatbot that says it is an AI, with a source for every answer.

Stefan shows you what the AI notice in the RAGSuite widget looks like and where you adjust it.