AI labelling requirement from August 2026: What companies need to know now

What does the AI labeling requirement regulate from August 2026?

As of 2 August 2026, the transparency obligations under Article 50 of the EU AI Act will apply in the EU. They require companies to clearly label and disclose certain AI-generated or AI-manipulated content, and to disclose when users are communicating with a chatbot instead of a human. In Germany, compliance is monitored by the Federal Network Agency (Bundesnetzagentur). Violations can result in fines of up to 15 million euros or 3 percent of global annual turnover, whichever is higher. For small and medium-sized enterprises, the assessment of fines is given particular consideration.

What matters here is not the size of the company, but whether AI is used professionally. If you do not yet know where AI is used in your business, you should first clarify this with a free AI readiness check before the labeling requirement becomes an issue.

Who is affected?

The requirement applies to all companies that use AI systems professionally, regardless of industry or number of employees. It affects both providers and operators of AI systems. In practice, this means: as soon as a company uses AI to create images, videos, audio content or texts for the public, or operates a chatbot on its own website, the labeling requirement applies.

These contents must be labeled

Images, videos and audio files

AI-generated or AI-manipulated images, videos and audio files must be labeled if they look deceptively similar to real people, places, objects or events and could appear authentic — so-called deepfakes. Examples: an AI-generated photo of a building or product, a realistically appearing image of a person, or a synthetically generated interview.

Comics, cartoons, icons, illustrations, abstract motifs, as well as minor edits such as color correction or sharpening, do not require labeling — nor do contents that are obviously unrealistic.

Texts

For texts, the requirement only applies if they inform the public about matters of public interest — for example news, political information or opinion-forming content. Marketing texts usually do not need to be labeled.

Chatbots and AI assistants

Anyone communicating with a chatbot instead of a human must be able to recognize this right at the start of the conversation. A hidden notice or one that only appears later is not sufficient.

What should the labeling look like?

The EU AI Act does not prescribe a fixed wording. What matters is that the notice is clearly visible and understandable. In practice, simple, direct formulations have proven effective:

• "AI-generated image" – as a visible notice directly on or in the image.
• "This text was created with the support of AI" – at the beginning or end of a text.
• A notice at the start of the conversation that a chatbot is responding instead of a human.

Practical example: How a trades business handles labeling

A typical case: A trades business has AI-generated before-and-after images created for its website to show possible results, and also operates a chatbot for appointment scheduling. Both cases fall under the labeling requirement: the images receive a visible note, “AI-generated visualization,” and the chatbot clearly introduces itself as an AI assistant at the start of the conversation. The effort involved is low — the key is to go through all AI applications in use once, consistently, before August 2, 2026, rather than overlooking it.

Frequently asked questions about the AI labeling requirement

When does the AI labeling requirement take effect?

The transparency obligations under Article 50 of the EU AI Act apply throughout the EU from August 2, 2026.

Do small companies also have to label their AI content?

Yes. What matters is not the size of the company, but whether AI is used professionally. However, small and medium-sized enterprises are given special consideration when fines are assessed.

Do AI marketing texts have to be labeled?

Usually not. The labeling requirement for texts only applies to content that informs the public about matters of public interest, such as news or political posts.

What penalties apply in the event of violations?

Fines of up to 15 million euros or 3 percent of global annual revenue, whichever is higher. In Germany, compliance is monitored by the Bundesnetzagentur.

Does my chatbot have to identify itself as AI?

Yes. Users must be able to recognize right at the start of the interaction that they are speaking with an AI system and not a human.

Next step: Review AI use in the company

Before the labeling requirement becomes a risk, it is worth getting an overview of where AI is already being used in the company. Our BAFA-approved AI consulting supports the legally compliant introduction and documentation of AI applications. If you are still at the very beginning, the best place to start is the free AI readiness check.