Art. 50: AI Labeling Requirements in Detail

Article 50 of the EU AI Act regulates transparency obligations for certain AI systems. The core rule: users must know when they interact with AI or consume AI-generated content. The obligation applies from August 2, 2026, and covers: • AI systems that interact directly with people (e.g., chatbots) • Systems that generate synthetic content (text, image, audio, video) • Deepfakes and biometrically manipulated content

The labeling requirement covers: • AI-generated text: When published to inform the public • AI-generated images: All synthetic or manipulated images • AI-generated videos: Including deepfakes • AI-generated audio: Synthetic voices, music • Chatbot interactions: Users must know they're communicating with AI Exception: Obviously creative or satirical content may be exempt from the requirement.

The EU AI Act does not prescribe a specific form, but recommends: 1. Machine-readable marking: Metadata in content (e.g., C2PA standard) 2. Visible labeling: Clear notices for end users (e.g., "This text was created with AI") 3. Chatbot notices: Clear notification at the start of interaction 4. Watermarks: Recommended for images and videos Practical tip: Implement a standardized process for all departments that publish AI-generated content.

✓ Create an inventory of all AI-generated content ✓ Define labeling processes for each department ✓ Establish responsibilities (who reviews, who approves) ✓ Plan technical implementation (metadata, visible labels) ✓ Train employees ✓ Document labeling practices ✓ Regular review and adjustment