Marking of synthetic content
Obligation of the provider of an AI system that generates synthetic audio, image, video or text content: the outputs must be marked in a machine-readable format so that it is possible to detect that they have been artificially generated or manipulated.
It is the obligation that what an AI generates carries a mark a machine can read, not a label the user sees.
Which obligations it carries
A legal obligation of the provider of the system, applicable from 2 August 2026. For providers that placed their system on the market before that date, Article 111(4) — added by Regulation (EU) 2026/1744 — gives until 2 December 2026 to take the necessary measures. For the deployer, verifying that its provider complies is an internal methodological control: there is no legal mandate requiring it by this route.
What it is not
It is not an obligation of the deployer. The deployer’s obligation on generated content is another — Article 50(4), disclosing that deep-fake content has been artificially generated or manipulated — and neither the party nor the object coincide. Nor is it a visible seal, a watermark for the eye or a notice to the user: the wording says “machine-readable format”. And it is not an absolute obligation: it does not apply where the system performs an assistive function for standard editing, where it does not substantially alter the input data provided by the deployer or their semantics, or where it is authorised by law to detect, prevent, investigate or prosecute criminal offences.
The nuance almost nobody captures
The technical requirement is expressly bounded: the solutions must be effective, interoperable, robust and reliable “as far as this is technically feasible”, taking into account the costs of implementation and the state of the art. It is not an obligation of perfect result, but nor is it an excuse: what modulates the standard is the technical feasibility recognised in standards, not the convenience of whoever applies it.
Related terms
To find out more
- The Digital Omnibus is now in force: what changes in the AI Act and what stays the same
- Marking synthetic content: whose obligation it is, and how far the transitional period reaches
- The code of practice on AI content transparency gets the green light: what it is and what it isn't
- Do you have to label all AI-written text? The editorial review route
- The EU's official icons for labelling AI content: how to use them and what they don't prove