On August 2, the European artificial intelligence regulation started to apply in earnest. Since that date, over there you have to tell people when they’re talking to a machine, and you have to label AI-generated content dealing with matters of public interest. The heaviest obligations, those for high-risk systems, were postponed to 2027 and 2028.
That was reported everywhere. And if your company sells to Europe or is a subsidiary of a European parent company, it reaches you directly.
But if you only operate in Chile, you probably thought “this has nothing to do with me” and went on with your day. It’s an understandable mistake, and an expensive one.
What nobody is looking at
In Chile there’s an artificial intelligence bill going around in circles in the Senate. It’s in its second reading, with a risk-based approach similar to Europe’s, and it has been there for months. Everyone is following it.
Meanwhile, Law 21.719 (the personal data law, passed in December 2024) takes full effect on December 1 of this year.
That law doesn’t say “artificial intelligence” in its title. But it regulates automated profiling and algorithmic decisions. In other words: exactly what your chatbot, your audience segmenter, your lead scoring system and that tool that decides who gets which email all do.
And it comes with administrative fines of up to 20,000 UTM (Chile’s inflation-indexed monthly tax unit).
It’s not a future law. It’s a law that has passed with a date set, and there’s just over three months left.
Why we’re telling you if we’re not lawyers
Because we’re not, and this note isn’t legal advice. Go to a lawyer if you need legal advice.
We’re telling you because we’re the ones who build your chatbot, your segmented campaign and your email flow. And because part of this isn’t legal: it’s about communication, and that part is our job.
When a rule requires you to declare that there’s a machine behind something, it stops being a compliance issue and becomes a matter of tone. You can write “This system uses artificial intelligence technology in accordance with current regulations” and sound like a contract’s fine print. Or you can write “A bot is answering you. If you get stuck, type human and I’ll put you through to a real person.”
Both comply. Only one builds trust.
Three things for this week
Make the list. Where you use AI today, with which tool, for what, and with what data. On one sheet. Without that list you can’t comply with anything, because you don’t know what you have.
Disclose what’s already running. If a bot answers your messages, say so. If you generated an image with AI, say so. It’s the cheapest obligation to meet and the one that builds credibility with your partners fastest.
Check what you’re pasting into the model. If you copy your partners’ data into an AI chat so it can draft something for you, you’re processing personal data. That’s the point where most small businesses are going to get tangled up, and it’s easier to fix now than in December.
The bigger picture
We’re an AI-based lab. It’s in our interest for AI to be used more, not less.
And that’s exactly why we’re saying this: distrust kills adoption faster than any regulation. Transparency isn’t the cost of using AI. It’s what lets you keep using it.
Those who disclose first will be at ease in December. Those who wait to be forced will be drafting legal notices in a hurry and sounding exactly like fine print sounds.
Want us to review where you’re using AI and how to disclose it without sounding like a contract? Write to us.