Deepfakes under the Commission's guidelines: labelling will cover nearly all commercial AI content

Deepfakes under the Commission’s guidelines: labelling will cover nearly all commercial AI content

From 2 August 2026, Article 50(4) of the AI Act applies an obligation to disclose that content has been artificially generated or manipulated by AI. The AI Omnibus does not change this date.

The Commission’s guidelines of 8 May 2026 interpret this provision very broadly. A deepfake is any realistic content generated or modified by AI; image, audio or video. It does not need to depict real persons or objects, it is irrelevant whether anyone intended to mislead the recipient, and commercial content does not benefit from the exception for artistic works. Labelling is therefore required for, among others, product visualisations, synthetic voices in advertising and virtual models in campaigns.

The obligation applies not only to agencies and contractors, but in many cases also to those commissioning the content. Penalties: up to EUR 15 million or 3% of global annual turnover.

The Commission’s icons can be used for labelling: “AI GENERATED”, “AI MODIFIED” and an “AI” icon with a description. They are available free of charge on the Commission’s website (HERE). Their use is not mandatory. What is mandatory is the labelling of content itself.

Before 2 August 2026, companies should review the materials they use as well as their contracts with agencies and content providers with regard to disclosure of AI use.