From August 2026, new labelling requirements for AI-generated content will come into force in the EU. Texts, images or videos created wholly or partly using AI must in future be clearly labelled as such. The regulations are based on Article 50 of the AI Regulation.

A draft code of conduct on the transparency of AI-generated content from the European Commission already provides insights into its implementation:

Providers of AI systems will integrate functions for the machine-readable and visible labelling of content generated by their systems. This is to be achieved through a combination of machine-readable markers, such as watermarks or hashtags, alongside a standardised EU label. The labelling must indicate whether the content was generated entirely by AI or merely edited with AI support. Providers will include a prohibition on the removal of these labels in their terms and conditions of use. In addition, providers will make free online detectors available to automatically identify such content. However, it is not yet possible to assess how reliable the evidence provided by such labels will be in cases of suspected cheating in written examinations.

An article by Janine Horn.