Smashing Magazine: EU AI labelling rules take effect

The EU AI Act’s Article 50 transparency requirements came into force on August 2, 2026. On August 13, Vitaly Friedman published a detailed breakdown on Smashing Magazine covering what the rules mean for product teams and interaction designers. The short version: labelling AI-generated content is now a legal obligation for any company with EU users, regardless of where that company is based.

The regulation requires disclosure in four categories. Deepfakes — images, audio, or video synthetically depicting real people or places in ways that could mislead — must be marked. Chatbots and AI agents must identify themselves as non-human when interacting with users. Text on matters of public interest that was produced by AI without a named human with editorial responsibility also falls under the rule. Emotion recognition and biometric categorisation tools must disclose their function.

For product designers, the most immediate implication is that the generic sparkle icon (✨) used by many interfaces to signal AI features is not a valid disclosure method. The EU distinguishes between labelling “AI-powered” features — which the sparkle approximately communicates — and explicitly marking AI-generated content, which requires a clear, persistent, and unambiguous signal. The EU Commission provides an official icon set for this purpose.

Human editorial involvement changes the equation. AI-generated text that has been substantively reviewed and edited by a named human with editorial responsibility may not require disclosure. Minor corrections — fixing grammar, adjusting punctuation — do not qualify as substantive editing. Design teams need to think through where editorial ownership in their content workflows actually lies.

The geographic scope is similar to the European Accessibility Act: it covers any company worldwide whose products reach EU users. Fines sit at roughly $20,000 per violation, with a one-year grace period before enforcement begins. For now, that gives product teams time to audit where AI generation appears in their interfaces and what disclosure patterns are currently in place.