EU Starts Requiring Chatbots to Disclose They're AI — Fines Reach 3% of Global Revenue

Entercast Consulting·

Starting August 2, the European Commission began actively enforcing the AI Act's transparency rules (Article 50) — the first phase of Europe's AI regulation with real enforcement, not just guidance. Chatbots now must explicitly disclose that the user is talking to AI, not a person.

What Changed

According to the European Commission's official site and law firms including Travers Smith and Greenberg Traurig, the new obligations include: chatbots and interactive AI systems must disclose they aren't human; deepfake content (image, video, or audio edited or generated by AI) must be labeled; AI-generated or altered content must carry machine-readable marks to make detection easier; and emotion recognition or biometric categorization systems (like facial recognition) must notify the people exposed to them. Enforcement falls to the AI Office and each member state's authorities, with fines reaching €15 million or 3% of a company's global revenue — whichever is higher.

Why It Matters

This is no longer a statement of intent — it's enforcement with real fines, starting today. Any company operating a chatbot, generating synthetic content (text, image, audio, video), or using facial/emotion recognition with any exposure to EU users needs to be compliant now, not at some future point.

The Impact for Brazil

PL 2338, still moving through Brazil's Chamber of Deputies, was designed closely following Europe's risk-based regulatory model — meaning similar transparency obligations (chatbots identifying themselves, labeling synthetic content) are a fairly likely path for Brazil too. Brazilian companies serving EU customers already need to comply now; those without European operations get, with this date, a realistic preview of what AI enforcement will likely require once Brazil's own AI bill is finally enacted.

Entercast's Take

AI transparency is moving from good UX practice to a legal obligation with real fines. Companies that already treat AI-use disclosure, synthetic content labeling, and biometric recognition notices as part of product design — not something bolted on later — won't need to scramble when this same requirement reaches Brazil.