For years the most common question about AI-generated content was: can you tell? That question is changing. As synthetic images, audio and video become more convincing, detecting them through visible mistakes is no longer a reliable strategy. Transparency has to move from the viewer’s eye into the structure of the content.
From August 2, 2026, Article 50 transparency obligations under the European AI Act apply in a range of cases involving AI-generated or manipulated content. Commission guidelines and the related Code of Practice try to translate the principle into procedures: machine-readable marking, labels where required and information people can understand.
This does not mean every automated intervention needs a giant warning. Denoising or a small assisted retouch is not equivalent to a fully synthetic face or a deepfake. The challenge is proportional disclosure: make visible what can substantially change the perceived origin, identity or event represented.
This is where provenance matters. C2PA and Content Credentials can attach verifiable information to a file about its history: which device or software created it, which transformations were made and which actors signed specific claims. They do not certify that an image is true; they make its path more legible.
For cinema, journalism, advertising and digital art this can directly change workflows. AI decisions should not be reconstructed at the end of a project but documented while production is happening: which scenes contain generation, which voices are synthetic, which images come from archives and which elements were reconstructed.
Credits may change too. They have traditionally identified who participated in a production. In the future they may also describe the technical path of a work—not to turn a film into a manual, but to distinguish what was photographed, performed, generated, restored or modified.
Transparency does not necessarily reduce the creative value of AI. It can do the opposite: when the process is readable, audiences can discuss the choice instead of spending energy trying to discover whether they were deceived.
DECLARED ORIGIN means this: the more invisible artificial intelligence becomes in the final result, the more visible the responsibility of those using it must become.