CSOAI
Initializing...
CSOAI
Initializing...
Machine-readable marking of AI-generated content becomes mandatory for new systems on 2 August 2026 at 00:00 CEST. Countdown, scope check and next steps below.
Time remaining
Three obligations, one hard cliff. The Code of Practice expects at least two active layers of machine-readable marking.
Providers must ensure AI-generated images, audio, video and text carry technical provenance data such as C2PA manifests and invisible watermarks.
Deployers must tell natural persons when they are exposed to AI-generated content. A small icon in the corner is not enough.
Manipulated content resembling real people, places or events, plus emotion recognition systems, need explicit, accessible labels.
Article 50 applies to providers and deployers of generative AI systems, chatbots, deepfakes and certain biometric systems.
Select your industry and AI system type for a fast, non-legal indication of whether Article 50 likely applies.
This is a simplified heuristic, not legal advice. For a definitive assessment, use the scope scorecard or book a readiness call.
Article 99 enforcement begins 2 August 2026 for new systems. Non-compliance can be per infringement and cumulative across outputs.