What is Prohibited AI?

Prohibited AI refers to AI uses that are completely forbidden in the EU. Where other AI systems may be used with varying requirements, the EU's AI Act draws a clear line for certain uses and says: you simply may not do this. It is the law's sharpest category.

The ban has applied since February 2025 and targets uses considered to threaten fundamental rights. Examples include so-called social scoring (rating people by behaviour, like a society-wide grading system), AI that exploits people's vulnerabilities to manipulate them, and biometric mass categorisation of people.

The idea is to set an ethical and legal floor: no matter how useful an AI solution seems, it must never be built this way. That protects both citizens and businesses from ending up on the wrong side of the line.

Why does prohibited AI matter for your business?

Most businesses have no intention of building anything prohibited — but it can creep in unintentionally, for example in how you collect and categorise data about customers or employees. Knowing the boundaries helps you avoid accidentally crossing them.

Knowing what is completely forbidden is also reassuring when you evaluate ready-made AI tools from suppliers. You can ask the right questions and avoid solutions that may become illegal to use.

Prohibited AI in practice

Say a supplier offers a tool that promises to automatically "read" customers' emotions via camera and sort them into groups for targeted influence. Depending on how it is done, such emotion manipulation and biometric categorisation can fall within the prohibited area — and then you should stay away, however attractive the offer sounds.

In practice, for most businesses it comes down to a simple checklist: know which uses are forbidden, and check new AI ideas and purchases against it before you start. If you are unsure, it is worth getting help to assess where the line runs.