Signal EU Makes AI Content Labels Mandatory for Realistic Synthetic Media Starting August 2

Summary

Under EU AI Act rules taking effect August 2, 2026, companies that create or deploy AI must label content that looks authentic but is not, marking the law's shift from principle to enforcement. Synthetic text, images, video, and audio designed to look truthful must carry a visible label and a machine-readable digital watermark; texts on matters of public interest must be labeled as AI-made if there was no human editorial oversight. Providers building generative systems must embed machine-readable markings, while deployers — those who publish AI-generated deepfakes or public-interest text — must disclose the AI origin clearly and promptly, including at the moment a user begins interacting with a chatbot. The obligation applies regardless of intent to deceive, and while pre-existing AI content and personal or clearly artistic/satirical/fictional works are exempted, new AI systems must comply from August 2 with a four-month grace period for existing ones. Noncompliance can bring fines up to 3% of global revenue; industry groups including the CCIA argue the deepfake definition has been interpreted too broadly, risking over-labeling of ordinary commercial content, a concern Meta echoed in a July 28, 2026 statement.