From 2 August 2026, the AI Act requires all AI-generated content, deepfakes included, to be marked and labelled. Every company that publishes a text, an image or a video produced by AI is affected, whatever its size.
The facts
- The obligation stems from Article 50 of the AI Act, applicable throughout the European Union from 2 August 2026.
- A code of practice, published on 10 June 2026 and currently being assessed by the European Commission, allows companies to demonstrate compliance without an individual application to the authorities.
- Two distinct strands: technical marking of the content on the model providers' side, visible labelling on the side of the companies that distribute that content.
- Signing up to the code remains voluntary. The legal obligations of Article 50 apply regardless.
Who is affected and how should you prepare?
The point that surprises companies most: the obligation covers every organisation that distributes generated content, well beyond the publishers of AI models: the campaign visual produced by AI, the presentation video, the text published on the website. Marketing and communication teams are therefore on the front line, even though they are rarely involved in compliance projects.
Preparation comes down to three reasonable steps: take stock of the AI-generated content the company publishes (and the content its suppliers produce for it), define a clear and uniform labelling rule, and document the process, the code of practice published on 10 June offering a framework for demonstrating compliance without an individual application. Dealt with now, the matter is settled in a few weeks; discovered in August, it becomes an emergency project.
AIxH's view
This is a point our AI maturity audits now check systematically, before the deadline turns into a compliance issue handled in a hurry. If your communication already uses AI, and it probably does, you are affected; the only open question is whether you are ready.