The European Union wants clarity on artificial intelligence, and it has confirmed a new set of rules governing how the technology is used. Starting August 2, deepfakes and other AI-generated content will have to be labeled in the name of greater transparency. The requirement is part of the AI Act, the EU’s landmark artificial-intelligence law, which is being rolled out in stages over the coming months, as planned.
What does this mean in practice? Beginning Sunday, companies will be required to state clearly and explicitly when users are interacting with AI, chatbots included. At the same time, any image or text generated by AI will have to carry a label visible to everyone. That can be done through watermarks or other detection markers.
How the new EU law works
An EU official, quoted by the French daily Le Monde, laid out the reasoning behind the decision.
«Generative artificial intelligence makes it possible to create disinformation on an unprecedented scale. Today it is increasingly difficult to distinguish what is real from what is not.»
That is precisely why the new AI Act requirement kicks in: to keep people informed and to preserve their ability to trust what they see, hear, or read. Some problems have already surfaced, however—chief among them that the new rules apply only to professional use, which means anyone using AI for personal reasons is exempt.
Content created to inform the public on matters of general interest must also be labeled if there is no actual person who can review and approve it. Exemptions are carved out for artistic, creative, satirical, and fictional works.
The first criticisms of the rules
The move has generally been welcomed, but it has not escaped criticism. It starts with Karen Massin of Google, who argues that the rollout could prove complex and counterproductive, sowing confusion among the very people the rules are meant to protect.
«If online content is flooded with AI labels and overlapping legal disclosures, it becomes harder for users to get the clear context they need,» she said.
Many observers are skeptical about how the law will actually be enforced, especially given how rarely regulation manages to keep pace with technology these days. «Better late than never,» some experts counter. It remains to be seen whether companies and content creators will adopt AI-content labeling right away, or whether the AI Act’s first penalties will start to bite. A first read on how it is going could come as early as next week.
Editor’s note
This article was originally published in Italian on money.it by Pasquale Conte on August 01, 2026 as «Usi l’AI? Ora l’Unione Europea vuole che lo dichiari chiaramente. Ecco in che modo». It has been translated and adapted for an international audience by the Money.it International desk.
