Effective Today: EU AI Act Transparency Rules Now Enforceable, Fines Up to EUR 15 Million
Starting today, the world's toughest AI regulation has grown teeth — the EU AI Act's transparency provisions are now enforceable, with fines up to EUR 15 million or 3% of global annual revenue.
Three Key Takeaways
Four transparency obligations take effect simultaneously. The European Commission's AI Office and national regulators will enforce four transparency obligations from August 2: first, AI systems that interact directly with users (such as chatbots) must clearly disclose that the user is talking to AI, not a human; second, AI-generated or modified content must carry machine-readable marks for detection; third, deepfakes (AI-edited or generated images, video, and audio) must be labeled; fourth, emotion recognition and biometric categorization systems must notify individuals being processed.
Fines rival GDPR levels. Violations carry penalties of up to EUR 15 million or 3% of global annual turnover, whichever is higher. Enforcement authority is split across three tiers: national market surveillance authorities handle general AI systems, the AI Office directly supervises general-purpose AI model providers (meaning OpenAI, Google, Meta, and other large model companies answer directly to Brussels), and the European Data Protection Supervisor oversees AI used by EU institutions themselves.
Content watermarking is now a legal requirement. Previously, AI content labeling was voluntary at best. Now it is a legal obligation. Every AI-generated image, every AI-synthesized audio clip must embed machine-readable metadata. The EU has simultaneously published a Code of Practice on Transparency of AI-generated Content to guide technical implementation.
WangDou's Take
Yesterday the U.S. handed the NSA classified authority to rate frontier AI models. Today the EU is waving the fine book. Two sides of the Atlantic said the same thing in completely different styles: the era of AI running naked is over. But the approaches diverge sharply — America is running security screening focused on how dangerous a model's capabilities are; Europe is running consumer protection focused on whether you told the user they are talking to a machine. The EUR 15 million cap sounds scary, but for companies pulling in tens of billions in annual revenue, it is the 3% turnover penalty that bites — Meta's EUR 1.2 billion GDPR fine from last year is still warm. From today, every AI chat window in Europe will carry an extra line at the top: "You are speaking with AI."
Source: European Commission Press Release, EU Digital Strategy