EU AI Act Transparency Rules

As of two days ago, chatbots operating in the EU are legally required to tell you they're not human. Not in the terms and conditions. In the actual conversation, the first time you talk to one.

Article 50 of the EU AI Act became enforceable on August 2, and it's narrower than people assume, which is exactly what makes it interesting. A watermark in the metadata doesn't satisfy it. A vague reference to "your assistant" doesn't satisfy it. The disclosure has to be perceivable in the interaction itself, to every new person who encounters the system, not just the first user ever.

The AI Office and national regulators also gained real enforcement power on the same day, not just monitoring authority. That's the part enterprises building or deploying conversational AI in Europe should sit with. This isn't a future compliance deadline anymore. It's active.

I'd bet this becomes a design default well outside the EU within 18 months, the same way GDPR's consent banners did. The companies that treat disclosure as a UX problem to solve well will be ahead of the ones treating it as a legal checkbox to survive.

If you're building AI products today, are you designing for this transparency requirement, or waiting to be told you have to?