New AI usage guidelines: A summary of the key facts for businesses.

New regulations: Since August, new, uniform guidelines for the use of artificial intelligence have been in effect across the EU. The so-called "AI Act" plays a key role here, especially for companies that use AI to create content. The goal of the new regulation is to promote greater transparency and maintain public trust.

Different risk classes and deadlines:‍

The new EU guidelines set out new requirements for companies using AI. Depending on the potential risk, AI usage is assigned to different categories. The basic rule is: the higher the risk, the stricter the guidelines.

Please note: AI-generated content that was published before August 2026 must also be labeled accordingly by the beginning of December 2026.

Who do the guidelines apply to?

The new guidelines affect both developers and companies that use artificial intelligence in their day-to-day operations. The country-of-origin principle applies: the regulations apply to all companies that use or offer AI systems in the EU, provided that EU citizens are affected by them.

Further information on the affected parties can be found on the official website of the IHK Frankfurt am Main .

Consequences of non-compliance:

Companies face significant fines for violating the requirements of the AI Act. Depending on the nature and severity of the violation, these can amount to several million euros or be based on a percentage of the company's total worldwide annual turnover. Precise details can be found

on the official website of the IHK for Munich and Upper Bavaria.

Labeling AI-generated text:

Texts must be labeled if the content relates to public interest topics, such as politics or news. Content that is manipulative in nature or strongly opinionated may also be subject to labeling requirements.

However, there are exceptions where the guidelines do not require disclosure. This primarily applies to texts where AI was used solely for support and the fundamental creation was performed by humans. Examples include spelling corrections or translations assisted by artificial intelligence.

How should the labeling be done?

In principle, a brief note can be added at the end of a text stating that the content was created with the help of AI. On websites, the disclosure can alternatively be placed in the legal notice (Impressum) or on a separate information page.

Labeling for visual content:

The use of AI has become widespread in recent years, particularly for images and videos. Artificial intelligence is increasingly being used to create visual content on social platforms and in advertising.

The AI Act aims to ensure greater transparency here as well.

A labeling requirement exists here, especially for AI-generated or manipulative content intended to mislead people—particularly in the case of deepfakes. This also applies to content in a political context, interviews, and, in general, visual material that has been created entirely from scratch and depicts real people or events in a deceptively realistic manner.

Similar to text, there is no immediate labeling requirement when AI is used solely for correcting or editing existing content. This includes, for example, adjustments to exposure and sharpness. The same applies to the removal of elements that do not alter the overall content.

Likewise, illustrations that can be clearly classified as unrealistic (such as comics or icons) and stylized advertising illustrations do not need to be identified as AI-generated.

How should the labeling be done?

If an image or video was created using AI, a corresponding notice should be included in the image or video description. For videos, the disclosure can also be made through clearly visible text overlays at the beginning or end of the clip.

For the labeling of visual content, the EU Commission has also royalty-free symbols published that companies can use for this purpose.

Labeling for AI interactions:

From now on, it must be clearly recognizable to users whenever they are interacting with artificial intelligence. This applies to AI chatbots on websites, automatically generated email responses, voice assistants in customer service, and AI-supported advisory systems.

A notice is generally not required if the interaction with AI is already clearly recognizable to users—for example, in applications explicitly labeled as AI. Furthermore, if the technology is used solely as a supporting tool for employees and the contact remains with a human, it does not necessarily need to be labeled as such.

How must the labeling be implemented?

As with text and image material, the use of AI should also be clarified during interactions. Notices can be provided at the beginning of a conversation, for example via audio, or as a note in automated messages.

Data protection and responsibility:

The new guidelines do not replace the General Data Protection Regulation (GDPR). If personal data is processed using artificial intelligence, existing data protection requirements continue to apply.

‍

Sources:

Artificial Intelligence Act – what companies need to know. (n.d.). IHK for Munich and Upper Bavaria. Retrieved August 5, 2026

Uniform rules for artificial intelligence in the EU. (2026). The Federal Government.

EU AI Act. (n.d.). IHK Frankfurt am Main. Retrieved August 5, 2026

EU icons for labeling AI-generated content. (2026). European Commission.

AI Regulation: New transparency requirements as of August 2026. (n.d.). Cologne Chamber of Industry and Commerce. Retrieved August 5, 2026