Your AI twin needs a label

A creator walked through the system he now uses for short videos: he uploaded about fifteen seconds of himself talking to the camera, got what he calls a very believable avatar, and plans to run his Instagram on autopilot at roughly eight dollars a video instead of a hundred for a human editor. The pipeline is clever and the numbers are real. What almost every tutorial like it leaves out is one line of law that has applied in the EU since 2 August 2026.

• The economics are genuine, and for small teams an avatar removes the hardest bottleneck in video, which is showing up on camera every week.
• The fix costs almost nothing: a visible label from the first second, for example "AI-generated avatar", repeated in the caption.

• The AI Act calls it a deep fake when AI-generated or manipulated video "resembles existing persons" and "would falsely appear to a person to be authentic or truthful". An avatar of you, saying words you never recorded, framed as you talking to camera, sits very close to that definition. It does not matter that the person is you or that the intent is harmless.
• Article 50 says the deployer "shall disclose that the content has been artificially generated or manipulated", "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure". A lighter version exists for evidently artistic, satirical or fictional work; a marketing reel usually is not that. This year's amendments push several AI Act deadlines to 2027 and 2028, but not this one.

Source: Κανονισμός (ΕΕ) 2024/1689 (AI Act), άρθρα 3(60), 50(4)-(5) και 113, σε αντιπαραβολή με τη ροή εργασίας avatar ενός δημιουργού (YouTube, 27/9/2026)