Article 50: marking AI content and deepfake disclosure

Article 50 of the AI Act mandates that AI-generated content, including text, images, audio, and video, must be clearly marked so users know it is not human-made. Deepfakes—realistic but fake audio or video—require disclosure unless the content is artistic, satirical, or fictional. A concrete contrast: Article 50 is about labelling the output itself, whereas Article 13 covers transparency about the AI system’s capabilities and limitations, and Article 52 deals with informing users they are interacting with an AI, not just marking content.

To avoid traps, remember that Article 50 applies to all deployers, not just providers, and the label must be detectable by a typical user, not hidden in metadata. A quick elimination trick: if a question mentions informing users about system risks or data use, it points to Articles 13 or 52, not 50. Also, note that deepfake disclosure is not required if the content is clearly fictional or artistic—this is a common distractor in exams.

To test yourself, ask: 'Does this scenario involve marking the output itself, or telling the user about the system?' If it is about the output, Article 50 is your answer. A memory aid: think '50 = output label', like a '50%' sticker on a product, not the manual inside the box.