The European Union is setting a new precedent for digital content transparency with its AI Act, specifically Article 50, which introduces significant obligations for labeling AI-generated content. Most of these provisions are slated to take effect on August 2, 2026, demanding greater clarity in an increasingly automated media landscape.
For online media publishing text, a key requirement will be to disclose whether text is AI-generated or manipulated, particularly when it aims to inform the public on matters of public interest. This applies broadly, ensuring that audiences are aware of the origins of the information they consume. Beyond text, the regulations cast a wide net over deepfakes. Deepfakes, defined as AI-generated or manipulated image, audio, or video content that either resembles existing persons, objects, places, or events, or creates realistic AI-generated representations that falsely appear authentic, will always require explicit labeling.
Compliance hinges on clear and accessible disclosure. Information regarding AI usage must be transparent, highly visible, and presented at the very first interaction or exposure to the content. This commitment to accessibility ensures that diverse audiences can readily understand the nature of the content. Furthermore, providers of generative AI systems face a technical obligation: their outputs must be machine-readable, allowing for automatic detection as AI-generated material.
However, the regulations also carve out important exceptions. Content does not require AI labeling if it has undergone substantial human review or editorial control, and a physical or legal person assumes editorial responsibility for its publication. For content that is distinctly artistic, creative, satirical, or fictional, the disclosure obligation is reduced. In these cases, it merely needs to be stated that the content is AI-generated in a way that does not diminish the enjoyment of the work. Interestingly, content that is clearly fantastical or physically impossible—such as depictions of dragons—is specifically exempt from the deepfake definition, acknowledging its inherent non-authenticity.
Additional exemptions include the use of AI systems for law enforcement purposes, such as crime investigation, which is not subject to these disclosure requirements. A critical date to remember is August 2, 2026. Content generated prior to this date does not require retroactive labeling. However, any text generated before this date but published on or after August 2, 2026, must be labeled, unless it falls under the scope of editorial control and human review. To aid in compliance, a code of conduct for transparency has been developed, alongside official EU icons for labeling. While the icons are optional, the act of labeling itself is not.