Every AI Chatbot Used in Europe Must Now Say It Is Not Human

Two days ago, on August 2, a new rule switched on across Europe. Any AI chatbot or automated system that talks to people now has to say, clearly, that it is not a human. Videos, images, and audio made with AI have to carry a visible label. The content also needs a hidden, machine-readable mark so software can detect it later. Europe just made AI disclosure the law.

This is not just a European problem. The EU AI Act has extraterritorial reach — meaning if any of your users are in the EU, the rule applies to you, whether your company is in San Francisco, Shanghai, or Amsterdam. Over 180 organizations have already signed a Code of Practice on AI content transparency. The EU's AI Office and national authorities are now empowered to investigate, request documents, and impose fines.

The fines are serious. Companies that break the transparency rules face penalties of up to €15 million or 3% of their worldwide revenue, whichever is larger. For banned practices — like AI that manipulates people without their knowledge — the ceiling jumps to €35 million or 7% of global revenue. Fewer than one in ten companies have even started listing their AI systems, let alone checking if they comply.

If you are building anything with an AI chat interface — a support bot, a sales assistant, an onboarding tool — and any of your users might be in Europe, you need a disclosure. A clear 'You are talking to an AI' message at the start of a conversation is the baseline. AI-generated images, video, or audio shown to users need a visible label. The time to build this in is now — before an enforcement letter arrives.