MarckDev
All articles

August 4, 2026 · 3 min read

The EU AI Act: what changes for businesses (and what doesn't)

The EU AI Act: what changes for businesses (and what doesn't)

The EU AI Act scares people more than it should. If you run a business and use some artificial intelligence tool — a chatbot, a text generator, a system that analyzes data — the first question is always the same: what do I have to do now to be compliant? The answer, for most companies, is simpler than it seems.

Let's look at what the AI Act is, how it works and what concretely concerns you, without alarmism.

What the AI Act is, in two minutes

The AI Act is the European regulation on artificial intelligence, the first of its kind in the world. It came into force in August 2024 and applies in phases over the following months and years. The underlying idea is simple: not all artificial intelligence is the same, so the rules change according to the risk a system poses to people.

It applies to those who develop these systems, but also to those who use them in a business. And it's precisely the "those who use them" part that concerns most Italian companies.

The risk-tier system: where you fit

The AI Act divides uses of artificial intelligence into four tiers:

  • Unacceptable risk (banned): practices such as social scoring of citizens or manipulation of behavior. These are prohibited, full stop.
  • High risk: systems used in sensitive areas — staff recruitment, credit assessment, medical devices, critical infrastructure. Here the obligations are serious.
  • Limited risk: tools such as chatbots or AI-generated content. The main obligation is transparency.
  • Minimal risk: the vast majority of everyday uses — spam filters, product suggestions, text correctors. No particular obligation.

The good news: most SMEs use AI in ways that fall under minimal or limited risk. High risk concerns specific sectors and uses, not "the company that put a chatbot on its website".

What you have to do in practice

If your business uses artificial intelligence in common ways, the concrete requirements are few and reasonable:

  • If you use a chatbot or an automated assistant: say so clearly to the user. The person has to know they're talking to an AI, not a human.
  • If you publish AI-generated text, images or video: they must be flagged as such, especially when they could be mistaken for real.
  • If you develop or use a high-risk system: here you need documentation, control of data quality and, above all, human oversight. This is the case for software that makes sensitive decisions about people.
  • Staff training: those who use these tools should have a minimum of AI "literacy" — understanding what it does and what its limits are.

In the projects we handle, when we integrate artificial intelligence features into a business system or an app, transparency and human oversight are part of the design from the start, not a patch added afterwards.

The deadlines that matter

The AI Act didn't all take effect at once. The banned practices are the first to apply, followed by the rules for general-purpose artificial intelligence models, and finally by the obligations for high-risk systems, which come fully into force between 2026 and 2027. If your use of AI is minimal risk, these dates barely touch you; if you fall under high risk, it's worth getting expert guidance in good time.

The fines (and the real risk for you)

The fines for the most serious violations reach up to tens of millions of euros or a percentage of worldwide turnover. These are figures designed for major violations, not for the small business that forgot to write "virtual assistant" next to its chatbot. For an SME using AI in an ordinary way, the practical risk is low: what matters is being transparent and knowing what you're using.

Want to use AI in your business and stay compliant?

The AI Act isn't a reason to give up on artificial intelligence, but an invitation to use it with judgment. In the custom software we develop we integrate AI features — like the built-in chat of CareCloud — with attention to transparency and human control. If you want to understand how to make the most of AI in your company without nasty surprises, let's talk it over in a free call: together we'll assess what you need and what the regulation means for you.

Related articles