Author: Kurt Schauer
Do we really have to label everything now?
Artificial intelligence saves time, is fun, and provides great ideas. However, to ensure that everyone on the internet can still recognize what is real and what comes from AI, the EU is introducing a labeling requirement starting August 2, 2026. This sounds stricter than it is at first. Because as long as you remain in control and only use AI as a clever helper, almost everything stays the same. We show you at a glance where the boundaries lie and how to avoid warnings.
Do you have questions about the use of AI?
The KAOS advertising agency team will be happy to assist you.
Enquire now without obligation!
When do you need to label?
The golden rule of the EU is: No one should be unknowingly deceived. If content appears to be real but was created by AI, a notice must be included.
You MUST label this
- Deceptively realistic photos and videos (Deepfakes): Do you want an AI to create an image or video that looks like a real recording? For example, a photorealistic image of a person in your clothing or a prominent person in a fictional situation. Since viewers might mistake this for a real photo, a corresponding notice is mandatory here.
- Artificially generated voices (Voice Cloning): An AI tool generates a synthetic, human-sounding audio track from text. Listeners must immediately be informed that this voice was machine-generated.
- Texts without human review: An AI-generated article on a public topic is copied directly from the tool and published without the text being corrected by a human, facts being checked, or editorial responsibility being assumed.
No labeling is required here
- AI is only your assistant: You have ChatGPT generate a text draft, but you rewrite it yourself, add your own thoughts, and check the facts. Because you bear full editorial responsibility, the labeling requirement does not apply.
- Art & obvious fiction: An AI calculates a comic graphic or a surrealist artwork (like a flying car over your company). Since it is immediately recognizable that it is not a real recording, there is no risk of deception – and thus no labeling requirement.
- Pure auxiliary tools: You only use AI to correct typos, rephrase sentences, or translate a text you wrote into another language.
- Everything that remains internal: Drafts, emails to colleagues, or presentations that will never be publicly available online.
The EU Commission’s Solution
If you need to label content, use the official icons of the European Commission. These EU symbols are uniformly designed and free for anyone to use.
The 3 Official Icon Types:
Base Symbol (AI):
Indicates that AI was involved in the creation process.
Completely AI (AI GENERATED):
For texts, images, or videos that were created entirely by AI without human refinement.
Partially AI (AI MODIFIED):
When genuine, human-created material has been subsequently altered by AI (e.g., deepfakes or photo editing).
Important Rules for Integrating the Icons:
- Early Visibility: The icon must be clearly visible at the latest when the user first encounters the content.
- Permanent Embedding: The symbol should be embedded directly into the image or video so that it remains intact when shared or downloaded.
- Consider Accessibility: Symbols must be provided with alt texts for screen readers (e.g., Alt=“Content artificially generated“) and have a legible font size.
The Underestimated Risk in Copyright and Contracts with AI
A central point that goes beyond the labeling requirement is copyright: pure AI outputs are in the public domain. German copyright law only protects personal intellectual creations by humans.
- No protection for pure AI images/texts: Those who use pure AI results do not acquire exclusive copyright or usage rights to them.
- Risk of legal infringements: AI models can reproduce third-party brands, faces, or copyrighted works. In the event of legal infringements, you as the operator are liable – often even without your own fault.
Safely adapt business processes
- Create internal guidelines: Define how AI may be used within the team and where editorial responsibilities apply.
- Utilize AI as an assistant: Always retain final editorial control – this saves labeling effort and protects your brand identity.
- Review software providers: When choosing tool providers, pay attention to data protection and indemnification clauses in case of legal infringements.
- Adapt contracts & approvals: Inform customers transparently about the use of AI tools and clearly regulate usage rights.
Our View on the New Transparency Rules
The new AI labeling requirement from August 2, 2026, is not an obstacle to modern marketing.
On the contrary: it rewards genuine content craftsmanship. As long as creative leadership and editorial responsibility remain with humans, your brand communication will not only stay unique but also largely free from labeling requirements.
Do you need support with implementation?
The team at the advertising agency KAOS is here to assist you with advice and practical help.
Simply contact us!
Related articles
Becoming Visible Online: A Guide for Your Business
A company's digital presence is a decisive factor for success. Many entrepreneurs invest in an appealing website, provide...
Social media for companies
Social media is essential for reaching customers and presenting yourself as an attractive and beneficial company on well-known platforms.
