01 What Article 50 requires
- Art.50(1)Providers
A system intended to interact directly with people — a chatbot, a virtual assistant, a customer-service bot — must be designed so those people are told they are dealing with AI, unless it is obvious from the circumstances.
- Art.50(2)Providers
Synthetic audio, image, video or text generated by the system must be marked in a machine-readable way, so its origin can be detected.
- Art.50(3)Deployers
Emotion recognition and biometric categorisation systems must inform the people exposed to them.
- Art.50(4)Deployers
Deepfakes — content that would falsely appear authentic — must be disclosed as artificially generated or manipulated. The same limb covers AI-generated text published to inform the public on matters of public interest, unless a human took editorial responsibility for it.
02 What Article 50 does not require
The duty is per system and per use — which is why a summary that reads as “label everything” is wrong in both directions: it overburdens the internal tools that owe nothing, and it hides that a chatbot owes its disclosure today.
- Not requiredA label on everything an employee produces with an internal tool
- Not requiredA disclosure for a back-office model no person interacts with
- Not requiredAny Article 50 duty at all for a system that neither interacts with people nor generates or manipulates content
03 The one grace period, providers only
Article 111(4) gives systems already placed on the market before 2 August 2026 until 2 December 2026 to comply with the Article 50(2) marking duty. Deployer duties under Article 50(3) and 50(4) had no grandfathering — they have been fully applicable since 2 August 2026.
Transparency duties applicable · Art.50(3) and 50(4) with no grandfathering
ApplicableArt.50(2) marking · systems placed on the market before 2 August 2026 · Art.111(4)
Grace period ends