EU brings in mandatory labelling of AI content
From 2 August, transparency duties under the EU AI Act require that you know when you are talking to a bot and when you are watching a deepfake.
- In May the European Commission published draft transparency guidelines under Article 50 of the EU AI Act.
- The duties take effect on 2 August and require disclosure when you are talking to AI, plus labelling of deepfakes.
- For visual content, the Commission proposes a common mark with the capital letters AI.
- Fines can reach up to 3% of global annual turnover or 15 million euros.
Will you soon know when the image on your screen was made by a computer? Yes, and by law. In May the European Commission published draft transparency guidelines under Article 50 of the EU AI Act, and the duties take effect on 2 August.
The idea is as simple as it should be. When you chat with a bot, it must be clear that it is a bot and not a person. When you watch a clip or look at a picture that AI created or altered, that has to be marked. No more guessing.
What exactly changes in August?
Two things that touch everyday life online. The first is mandatory disclosure that the user is communicating with artificial intelligence, so chatbots on websites and in apps must present themselves as what they are. The second is the labelling of deepfakes, clips and images in which a face or a voice has been artificially created or altered.
The Commission also suggests how the mark should look. For visual content there is a common label with the capital letters AI, so it is recognisable at a glance. For long videos you do not just get it at the start and then forget it: the mark is shown at the beginning and then at regular intervals while you watch.
What if someone ignores the rules?
This is where it gets serious. Fines under these duties can reach up to 3% of a company’s global annual turnover or 15 million euros. For large platforms that is not small change you write off, so it is realistic to expect the marks to appear even before the first inspection arrives.
For you as a user the change is quiet but useful. You will spend less time wondering whether you are speaking to a live support agent and whether a viral clip is real. Our advice is to get into the habit of looking for that mark, the way you now check whether a site is on https.
Will you trust what you see on screen more, now that AI content has to be labelled?
