New EU regulations will require clear labeling of AI-generated content starting August 2026. Publishers and businesses must identify which materials need disclosure and understand the risks of non-compliance.
Starting August 2, 2026, the EU’s Article 50 AI Regulation will require anyone using AI systems professionally to clearly label certain AI-generated or AI-altered content. This move directly impacts publishers, marketers, and businesses that rely on generative AI for commercial purposes, making it essential to understand which materials fall under the new rules and where compliance risks are highest.
The regulation’s scope extends beyond major AI providers like OpenAI or Anthropic. According to legal expert Patrick Schulz, the obligations also apply to any individual or company using AI systems independently for business. This means nearly all commercial users of generative AI must pay attention to the new transparency requirements.
Labeling is especially critical for AI-generated or manipulated images, audio, and video-commonly known as “Deepfakes.” The rules are less strict for AI-generated text, except when the text is published to inform the public about matters of public interest. Standard marketing copy, product descriptions, newsletters, and most social media posts are generally exempt, especially if a human reviews the content before publication and takes responsibility for it.
The regulation defines Deepfakes as AI-created or altered visual, audio, or video content that could be mistaken for real people, places, objects, or events. While the law’s wording suggests a resemblance to real entities is required, EU Commission draft guidelines interpret the definition more broadly. Even realistic depictions of non-existent people or places may require labeling if they could be perceived as authentic. The rules also cover genuine media that has been modified or enhanced by AI.
There are notable exceptions. Content that is obviously unrealistic-such as a person flying unaided or a talking mouse-does not require labeling. Routine AI-powered edits like color correction, brightness adjustment, noise reduction, or image scaling are also excluded from the requirements.
For businesses, the highest risk lies in publishing AI-generated content that intentionally mimics reality, such as photorealistic ads, product images, cloned celebrity voices, or virtual influencers. In contrast, clearly unrealistic visuals, standard edits, or human-reviewed texts carry much lower risk.
The law mandates that disclosures must be “clear and unambiguous” and meet accessibility standards. While no specific format is required, options include text labels like “AI-generated,” visual icons, or audio cues. For German audiences, using German-language labels is recommended, as English-only tags or hashtags like #AIgenerated may not meet legal standards.
Companies deploying AI chatbots on websites or in customer support must also inform users that they are interacting with an AI system. This notice should appear at the start of the interaction, unless it is already obvious to the user. Providing a clear disclosure is a low-effort way to reduce legal risk.
Given the broad interpretation of Deepfakes, experts advise erring on the side of transparency for any realistic content created or altered by AI. Failing to comply can result in fines and legal claims from competitors or industry groups. For more on how publishers are navigating AI’s impact on their operations and revenue, see this analysis of how AI is reshaping open web funding and media buying.
