EU labelling duty under Article 50 reaches deep fakes, not all AI content

The European Commission's page on labelling AI-generated content, last updated 20 July 2026, states that "Not all AI-generated or manipulated content needs to be labelled." The disclosure duty under Article 50(4) of the AI Act reaches deep fakes, defined as content resembling existing persons, objects, places or events, and AI-generated text published on matters of public interest without human review.

On the face of that definition, a wholly AI-generated drama with invented characters, places and events does not resemble anything existing. Where a deep fake does appear in what the text calls an "evidently artistic, creative, satirical, fictional or analogous work", the obligation is limited to disclosure "in an appropriate manner that does not hamper the display or enjoyment of the work".

Four limits sit alongside that reading. The human-review carve-out applies to text only: a real likeness, a de-aged actor or a synthetic voice of a living performer is a deep fake and the duty applies. The Commission's guidelines are non-binding, and only the Court of Justice can interpret the Article authoritatively. Signatories to the Code of Practice remain bound by its Section 2 regardless. And limited disclosure is still disclosure.

A label in every episode is a requirement of China's Article 34, not of the EU text.