Four Days From Now, "You Are Talking to AI" Becomes the Law

4 days from now, "you are talking to AI" stops being good manners and becomes the law.

This Sunday, August 2, the transparency rules of the EU AI Act kick in. If your business serves European customers, here is the short version, no legalese.

What actually starts on Sunday:

1. Your bot must say it is a bot. Clearly, at the start, not in a footnote.
2. AI-generated content needs to be labeled as such.
3. Deepfakes and synthetic media: labeled, always.
4. Reading customers' emotions with AI: heavily restricted, and your people must be told if it touches them.

What got postponed, so you do not panic about the wrong thing: for systems already on the market, the content-labeling duty moved to December 2. A four-month breather, for that one piece only.

The fines go up to 15 million euros or 3 percent of global turnover.

But honestly, the fine is not the interesting part. The interesting part is that the law now demands the bare minimum your customers wanted all along: to know whom they are talking to.

Which means the companies that treat Sunday as a checkbox will do the minimum and move on.

And the companies that treat it as an occasion will tell their customers openly: here is our AI, here is what it can do, here is how to reach a human.

Same legal requirement, completely different relationship coming out of it.

I wrote a free guide on doing exactly that, the caring way instead of the checkbox way: Your Company's AI Through the Customer's Eyes. Checklist inside, no email required.

Four days. Enough time to check your bot introduces itself. Maybe even enough to make it something your customers thank you for.