Europe's AI Systems Now Have to Introduce Themselves — the First Days of Article 50 in Practice

The EU AI Act's Article 50 transparency obligations, in force since August 2, make Europe the first market where an AI system is legally required to tell the humans it talks to that it is artificial. The headline consumer chatbots complied quickly — a disclosure line costs a frontier lab nothing. The friction, visible in the first days, sits in the long tail: white-label support bots, AI features embedded inside SaaS products, voice agents on phone lines, and AI-generated public-information content, where the operator deploying the system — not the model vendor — carries the obligation and often does not yet know it. Two design questions are emerging as the practical battleground. First, what counts as adequate disclosure: a one-time banner, a persistent label, or an answer the system must give when asked directly? The Commission's July 20 guidelines sketch expectations, but enforcement practice will draw the real lines. Second, marking AI-generated public content: text published to inform the public needs visible marking, which quietly covers far more marketing and editorial output than most publishers have acknowledged. The compliance asymmetry favors platforms that are honest by architecture. Systems that already identify agent traffic, label machine-generated content, and expose what was automated have nothing to retrofit; systems built to blur the line between human and machine now carry both engineering debt and legal exposure — up to 15 million euros or 3 percent of global turnover. Disclosure-by-design has abruptly moved from an ethics preference to the cheaper engineering option.