AI Act (from 2 August 2026)
- Deepfakes — realistic AI images, video or audio of people, places or events — must be disclosed as AI-generated
- AI-generated text published to inform the public on matters of public interest must be disclosed, unless a person reviewed it and takes editorial responsibility
- For evidently artistic, satirical or fictional works, a light disclosure that doesn’t spoil the work is enough
- Providers of AI tools must also mark outputs in a machine-readable way (watermarks, metadata)
Slovenian media: stricter
Under the Media Act (ZMed-1), applied since 27 September 2025, media content created fully or partly with generative AI must be clearly recognisable, separated from other content and labelled at its beginning and end. Media must also explain how they use AI. This goes further than the AI Act — it applies to all such content, not only deepfakes.
Platforms ask too
YouTube, TikTok, Instagram and Facebook ask creators to disclose realistic AI-generated or altered content when uploading, and may add a label themselves. Not disclosing can lead to removed posts or reduced reach.
How to label well
- Short and visible: “Image generated with AI” or “Text prepared with the help of AI, edited by [name]”
- Where it is seen: on the image, at the start of a video, next to the text — not hidden in the terms
- Honest about the role: a fully generated image differs from a photo with a retouched background
- Keep the original metadata (Content Credentials) instead of stripping it
When labelling isn’t needed
Ordinary editing help — spelling, translation you checked, cropping, noise removal — generally doesn’t need a label outside the media sector. When in doubt, a short honest note costs little and builds trust.