Unmarked synthetic content

What it is

When content can no longer be told apart from the real thing by looking, the audience depends on being told. The EU AI Act places duties on both sides: providers of generative systems must ensure that outputs are marked in a machine-readable way as artificially generated, and those who deploy systems to produce deepfakes, or text published to inform the public, must disclose that the content is generated, with limited exceptions. The risk is that an organisation generates and publishes content with neither marking nor disclosure. Telling a person that they are talking to a machine is a separate duty, covered in undisclosed AI interaction.

How it happens

• The generation tool applies no marking, or the marking is stripped in editing.
• Marketing uses generated images and treats them like stock photography.
• Nobody is responsible for deciding when disclosure is needed.
• Content is edited by a person and assumed to be exempt.

Why it matters

It is a breach where the duty applies. Beyond that, an audience that discovers undisclosed synthetic content tends to distrust the rest of what the organisation publishes.

Warning signs

• No policy states when generated content must be labelled.
• Tools in use have never been checked for marking.
• Published images or video cannot be traced to their origin.
• Export and editing steps remove metadata.