Entertainment companies and rights holders face new obligations to identify certain AI-generated or manipulated material under European Union rules which came into force on 2 August 2026.
The requirements form part of the EU AI Act and apply where content is considered a “deep fake” – a category that can extend well beyond the deliberately absurd or obviously artificial clips that have traditionally attracted public attention. Virtual presenters, synthetic extensions to sets and computer-generated graphics assisted by AI may all fall within the rules, depending on how realistic they appear and the impression they create for viewers.
The legislation is European, but its reach is not necessarily confined to companies based inside the bloc. A business anywhere in the world may be caught if it generates material with the prospect that it will be distributed in the EU, including through publication on the internet. That gives the rules particular relevance to international broadcasters, production companies, sports organisations and entertainment platforms working across borders.
When realism becomes the key test
The central question is whether AI-generated or altered material resembles an existing person, place, object, organisation or event and could wrongly be taken as authentic or truthful. The definition is deliberately broad: the subject does not need to be a real individual or location, provided it could plausibly exist.
“Would falsely appear to a person to be authentic or truthful.”
That test means the surrounding presentation matters. A realistic virtual host for a travel programme, for example, may be treated as a deep fake if the presenter appears to be a genuine person reporting from real places in a format audiences expect to be factual. The European Commission’s guidance expressly treats lifelike AI-created avatars and personas as persons for this purpose.
Examples likely to require disclosure
The guidance also considers a comedy programme showing real politicians carrying out genuine actions in a fabricated, photorealistic setting. Although the people and actions may be real, the artificial environment can create a misleading impression about what actually happened. The assessment is made with the foreseeable audience in mind, rather than simply asking whether the majority of viewers would recognise the trick.
- Virtual presenters: realistic AI-created hosts may need to be identified when presented as genuine people.
- Synthetic locations: a fabricated setting can alter how an otherwise real event is understood.
- AI-assisted graphics: production techniques that create realistic material may fall within the definition.
- Online distribution: publication on the web can be enough to bring international businesses within the Act’s reach.
A new production consideration
| Issue | Practical significance |
|---|---|
| Rules in force | Since 2 August 2026 |
| Potentially covered material | AI-generated or manipulated content resembling reality |
| Geographical reach | Businesses worldwide where EU dissemination is foreseeable |
| Central assessment | Whether audiences could mistake the material for authentic |
For producers, the rules introduce another consideration at the earliest stages of a project: not simply whether AI has been used, but how the finished material may be understood by its audience. A graphic or virtual performance that is plainly fantastical may raise different questions from one designed to look like an unaltered recording of a real person or event.
The Commission’s guidance is intended to explain how disclosure should work, but the available examples are described as limited and fairly obvious. That leaves the practical boundary between ordinary AI-assisted production and material requiring a label dependent on the realism of the result and the risk of false authenticity. The issue is especially sensitive in formats such as sport, news-style entertainment and factual programming, where viewers may reasonably expect what they see to represent events as they happened.
As AI becomes a routine production tool, the debate is therefore moving away from whether artificial techniques are being used at all. The more consequential question is whether a finished programme, image or performance could persuade someone that an invented or altered scene is genuine. For international entertainment businesses, that judgement now has to be made with the EU audience firmly in view.
