The EU Now Asks AI to Say So. Here Is What That Means
By Norvel M A Roosje ·
Chatbots have started being more upfront about what they are. If you have talked to one recently, you may have noticed it saying, right at the start, that you are talking to AI and not a person. This is not just good manners. As of August 2, 2026, it is a rule.
The European Union passed a law called the AI Act, and one part of it deals with something simple: honesty. People should know when they are talking to a computer instead of a human, and they should know when a photo, video, or piece of writing was made by AI instead of a person. The rule is called a transparency rule, and it applies across the European Union.
What the rule actually says
Strip away the legal language, and the rule comes down to three plain ideas.
First, if a business uses a chatbot, AI agent, or a virtual character on its website, it has to tell visitors this clearly, from the very first message. Not hidden in a menu. Not in small print. Right there, at the start of the conversation.
Second, if a business uses AI to create an image, video, or audio clip that looks like a real person, place, or event, and that content could be mistaken for something genuine, it has to be labeled as artificial. This matters most for what people call deepfakes: content made to look real when it is not.
Third, if AI writes text about a serious public topic, such as politics, public health, or safety, and nobody actually reviewed it, that text needs a label too. If a real person read it, checked it, and takes responsibility for it, no label is needed. The point is not to ban AI writing. The point is to stop AI writing from quietly pretending to be careful, human journalism when it is not.
This is not only a European problem
Here is the part many business owners outside Europe miss. The rule does not only apply to companies based in the EU. It also applies to businesses outside the EU if people in the EU end up seeing or using what that business’s AI system produces.
For a business in Curaçao and rest of the Dutch Caribbean, this is closer to home than it sounds. The Dutch Caribbean has deep, ongoing ties to the Netherlands, which is an EU country. Family visits, tourism, students, and business travel between the islands and the Netherlands are common, not rare. A local business website written in both English and Dutch is very likely to get real visitors from the EU, not just a hypothetical one on paper. Being based outside Europe does not automatically put a business outside the reach of this rule.
What happens if a business ignores it
For companies that fall clearly within the EU’s reach, the fines on paper are large: up to 15 million euros, or 3 percent of the company’s yearly worldwide income, whichever amount is higher. Regulators are also expected to go easier on small and medium businesses than on large ones.
But for most smaller businesses, especially those outside the EU, the fine is not really the risk that matters. The bigger risk is quieter and closer to home: a customer figuring out, on their own, that they were talking to a computer that never said so. Once someone feels a little tricked, even by something small, trust does not come back easily. And trust is exactly what a small business depends on more than a large one does.
A real example: CaribWare’s own chatbot
We build websites for businesses across the Dutch Caribbean, and we also build tools like chatbots into those websites. So when this rule came into effect, we did not just write about it. We looked at our own site.
The chatbot on the CaribWare site (coming soon, by the way!) greeted visitors with a simple line: “I’m CaribWare’s virtual assistant.” That sentence is true. It is also soft. A visitor reading it quickly could easily walk away thinking they were about to chat with a real staff member, or at least not think about the question at all.
We changed one word. The greeting now reads: “I’m CaribWare’s AI assistant.”
It looks like a small edit. In practice, it changes the whole first impression. “Virtual assistant” is a phrase that has been used for years to describe both human support staff working remotely and actual software. It leaves room for doubt. “AI assistant” leaves no room for doubt at all. A visitor knows, in the very first sentence, before they type a single question, exactly what they are talking to.
That is the entire point of this rule. Not to slow AI down, and not to punish businesses for using it. Just to make sure people are not left guessing about something this basic.
What this means for your own website
If you run a website for your business, here are a few plain, practical things worth checking.
If you have a chatbot or any kind of AI assistant on your site, make sure it says so clearly, right in its first message, not just in a small title bar that visitors might not even notice.
If you use AI-generated images anywhere on your site, especially photos of people or scenes that look realistic, ask yourself honestly whether a visitor might believe they are looking at a real photo of your actual team, office, or location. If the answer is yes, either use a real photo instead, or make it clear the image is illustrative.
If AI helps write content for your site, particularly on topics that matter to your readers’ health, money, or safety, have an actual person read it, check it, and stand behind it before it gets published. That single step both satisfies the rule and, more importantly, keeps your content trustworthy.
None of this requires expensive tools or legal advice. It mostly requires a short, honest look at your own website, the same kind of look we gave ours.
Curious whether your website measures up?
Contact CaribWare and we will take a look together. No pressure, no obligation, just an honest review of what is on your site today and what, if anything, is worth changing.
Written with AI assistance. Thorough review and editorial oversight by Norvel Roosje before publishing.
CaribWare builds professional, honest websites for businesses across the Dutch Caribbean.