An EU AI ad that flatters the product counts as a deepfake
The Commission's final Article 50 guidance is tougher on advertising than many expected.
Background
The European Commission published its final guidelines on Article 50 of the AI Act on 20 July 2026. For advertising, the guidelines resolve a question that had been open since the Act was adopted: how does deepfake disclosure apply to commercial content?
Three key points
1. Flattering AI images are deepfakes. The guidelines state that an AI-generated image of a real product shown against an AI background is acceptable if it does not mislead about the product. An AI image that makes the product look different from or better than reality is classified as a deepfake and must be labelled. The threshold turns on whether the image could deceive a consumer about what the product is or does.
2. The creative exemption is narrow. The reduced labelling regime for evidently creative or satirical work is read restrictively. The advertising examples given in the guidelines all fall outside it. Brands and agencies should assume the exemption does not apply unless specific legal advice confirms otherwise.
3. The final output is what matters. Intermediate outputs in closed production workflows do not need to be marked. The obligation attaches to the final asset that goes to market, which determines where in the workflow labels need to be applied.
Why this matters
The guidelines are not soft guidance: they represent the Commission's authoritative interpretation of binding rules already in force. Agencies and brands that have not reviewed their AI content workflows against these standards should do so before their next campaign goes live. The question is not whether the rules apply, but how.
Sources: Bird & Bird; European Commission; Paul Weiss.
