Label It or Pay: The EU’s Deepfake Disclosure Rules Go Live in August

On 2 August 2026 — just over four weeks away — the European Union becomes the first G7 jurisdiction to make labelling AI-generated content and deepfakes a binding legal obligation, backed by fines reaching into the millions. It is the most consequential transparency law the synthetic-media age has produced. It is also, on its own, unlikely to stop a single determined fraudster — and understanding why is the point.

What Happened

On 2 August 2026, the EU AI Act's Article 50 transparency obligations become enforceable across all 27 member states, imposing binding disclosure requirements on chatbots, deepfakes, and AI-generated content for the first time in any G7 jurisdiction. The obligation and the power to fine for breaching it arrive together on that date.

Article 50 imposes four core duties. Providers of AI systems that interact directly with people must design them so users know they are dealing with AI (Article 50(1)); providers of generative AI must ensure outputs are marked in a machine-readable format and detectable as artificially generated (Article 50(2)). A third duty covers notice for emotion-recognition and biometric systems, and a fourth — the most scrutinised — requires deployers to label deepfakes. Notably, content that looks or sounds like a real person must be labeled even if no deception was intended and even if no real individual is depicted.

The rules reach beyond Europe's borders. The Act applies to providers and deployers in third countries where the output of the system is used in the EU — a US or UK advertiser running an AI deepfake in an EU-facing campaign is squarely in scope. And the bar for "disclosure" is deliberately high: a reference in the terms and conditions is not sufficient, technical labels such as metadata or watermarks alone do not meet the requirement, and vague terms such as "assistant" or "this system uses LLMs" also fall short. Penalties reach €15 million or 3% of worldwide turnover.

Why It Matters

For years, the response to synthetic media has been voluntary: platforms labelling what they chose to, companies disclosing when they felt like it. Article 50 ends the voluntary era in the world's largest single market, and its extraterritorial reach means the effects will be felt far outside Europe.

For any business putting AI in front of people — a chatbot, an AI-narrated video, a synthetic marketing image — "say so, clearly" is about to become law rather than etiquette. That is a genuine shift in the information ecosystem. The harder question, which the headlines skip, is whether a labelling law can touch the actors who were never going to label anything.

advertising