From Sunday the EU Can Fine Companies Under the AI Act, and Deepfakes Must Be Labelled

From Sunday the European Union can enforce its artificial intelligence rules and fine companies that break them. The AI Act was adopted in 2024, but its provisions have been introduced in stages, and this is the one that carries penalties.

The law scales obligations to risk: the greater the threat a system poses to health or to rights, the more its operator has to do to protect people from harm. Some uses are prohibited outright, including predictive policing, emotion recognition in workplaces and schools, and systems that manipulate human behaviour.

Two duties will be visible to ordinary users. AI generated content such as deepfakes must be labelled, and systems like chatbots must make clear to the person using them that they are talking to a machine. Systems already on the market have until 2 December to adapt, and there are exemptions for artistic, creative, satirical and fictional work.

The Union already runs the world's most comprehensive digital rulebook across social media, online retail and search. What changes on Sunday is not the ambition but the consequence: a rulebook with fines behind it is the point at which compliance departments start reading it properly.