Source: France-Antilles (Guadeloupe.franceantilles.fr)

From 2 August, content created or manipulated by artificial intelligence must be clearly identifiable across the European Union as a new set of transparency obligations under the AI Act takes effect. The measures target deepfakes, hyperreal images, synthetic voices and interactive AI systems, and include published models of icons ("IA", "generated by AI", "modified by AI") and an online guide from Brussels to help companies comply; non‑compliance exposes providers to fines. Chatbots and public-facing AI interaction systems must be presented as such; providers are required to integrate digital watermarks to enable detection of AI-generated material; deepfakes and AI-generated texts, images, videos or sounds must carry labels when published in a professional context; and texts on matters of public interest produced by AI without human editorial control also require labelling. Existing AI systems have until 2 December to adapt.

Several major platforms have already implemented labelling: TikTok reports billions of items labelled via detection tools, Meta uses an "AI Info" tag on Instagram and Facebook, and Google — a signatory to an EU transparency code — is reportedly working with Nvidia, OpenAI and Apple on digital marking solutions. Exemptions cover artistic, satirical, fictional or creative works and minor editorial assistance such as spellchecking or reformulation. Stakeholders warn of a potential "complexity regulatory" risk that could overwhelm users if online content becomes saturated with AI labels; Brussels frames the move as balancing transparency and usability, with the rules' real-world effectiveness to depend on clarity and adoption by digital actors.