AI as a Tool
With the new possibilities offered by AI that’s always available and programs that keep getting better, creating content (e.g., for websites or social media) is becoming easier and easier. As a result, we’re seeing more and more specialty retailers use AI-generated text and/or videos to draw attention to their brick-and-mortar stores or new offerings. Perhaps you’ve even tried your hand at creating content yourself using an AI tool like ChatGPT, Gemini, or Claude.
Artificial intelligence is constantly evolving and getting better—so much so that it’s often impossible to tell whether content was created by a human or an AI. That’s why a new labeling requirement for AI-generated content went into effect this month (August 2026). To help you avoid penalties, we’ve summarized the key points of the new regulation and what steps you need to take.
Labeling Requirements Under Article 50 of the AI Regulation
The European regulation requiring the labeling of AI content applies to providers and operators—that is, those individuals and companies that develop an AI system or AI model, have one developed, or use one under their own responsibility. Another key factor in determining the labeling requirement is whether AI is used primarily or only as a support tool.
- Primary AI-generated content: The content is generated independently by AI, without human editing, and in such a way that it might give the impression of having been written exclusively by a human. = Mandatory labeling
- Supportive AI Use: The AI merely creates a draft. The final text is shaped by human editorial work, or a human writes the text and the AI only assists with phrasing, proofreading, or translation. = No disclosure requirement
- Non-public use: Content used exclusively internally or in a private setting does not need to be labeled.
Examples
Texts
Texts that must be labeled include news, political information, socially relevant topics, or manipulative, opinion-shaping texts.
Product descriptions, texts that have been improved by AI spell-checking, AI-generated translations, and, in general, texts that were primarily created by a human, reviewed by a human, and for which a human assumes editorial responsibility do not need to be labeled.
Images/Videos/Audio
The following must be labeled: AI-generated people, artificially created interviews, manipulated corporate videos, and product images that are not based on real photos, videos, or physical prototypes (e.g., an image of an unreleased iPhone model), AI-generated voices, AI-generated photos of a company building, or AI-generated images depicting an event that did not actually take place (e.g., an AI-generated image or video of you in a fictional customer meeting).
Illustrations (caricatures, comics, icons, abstract graphics), promotional illustrations, or images that have been edited only in terms of sharpness, exposure, or color—or from which insignificant elements have been removed (e.g., trees or people in the background)—do not need to be labeled. Provided they are not misleading, images, videos, and audio files also do not need to be labeled.
Here’s How to Label AI
- Texts subject to labeling requirements should include the statement “This text was created with the assistance of AI” (or similar) at the end.
- Images, videos, and audio files subject to labeling requirements should be accompanied by appropriate visible or audible notices, such as “AI-generated,” “This audio recording was created with the assistance of AI,” or “This video was created in whole or in part using artificial intelligence.”
- The EU has provided symbols that can be downloaded and inserted: https://digital-strategy.ec.europa.eu/de/policies/eu-icons-labelling-ai-generated-content
Further information:
- https://gesellschaft-datenschutz.de/ki-kennzeichnungspflicht/
- https://www.ihk.de/koeln/hauptnavigation/digitalisierung-und-innovation/digitalisierung/transparenzpflichten-nach-der-ki-verordnung-7100068
- https://digital-strategy.ec.europa.eu/de/policies/code-practice-ai-generated-content
