Artificial Intelligence

From 2 August 2026 the EU AI Act’s transparency chapter makes conversational and generative AI announce itself: chatbots must say they are chatbots, deepfakes must be labelled, and generated content must carry machine-readable marks. The interesting part is what that reveals. Eurostat puts generative AI use at 32.7% of Europeans aged 16 to 74, while a consumer study across six European markets found only about a third recognise using AI-powered services daily. This guide covers the four Article 50 duties and who owes each, what Europeans will visibly notice across banking, support lines, marketing and retail, and the much larger category of everyday AI – fraud scoring, ranking, dynamic pricing, routing, triage – that owes no disclosure at all. It also covers the Digital Omnibus deferral that pushed the high-risk rules for hiring, credit and education to December 2027, the four-month grace period on machine-readable marking, the EUR 15 million or 3% penalty exposure, and a four-step readiness sequence for organisations selling into the EU.