EU rules taking effect from Sunday will require companies to label AI-generated or manipulated images, audio and text that are designed to appear authentic. The measures, introduced under the EU AI Act, are intended to help people identify deepfakes and other synthetic material, with supporters arguing that greater transparency is important for consumer protection and democracy.

New AI systems placed on the EU market must comply from 2 August, while existing systems have four additional months. Content must carry a visible AI label and a digital watermark; public-interest text without human editorial oversight must also be marked, though personal content and clearly artistic, satirical or fictional works are exempt. Companies that breach the rules could face fines of up to €15m or 3% of global turnover, while technology industry groups warn that overly broad labelling could desensitise users.