Artificial Intelligence (AI) is changing the way we live, work, and conduct research. It can improve healthcare, support scientific discoveries, help protect the environment, and make businesses more efficient. However, not every use of AI is acceptable. Some AI applications create unacceptable risks to people, society, and fundamental rights.

This is why the EU AI Act identifies a number of prohibited AI practices. These are AI systems that cannot be developed, marketed, or used in the European Union because the risks are considered too serious.

For organisations involved in Horizon Europe projects, research institutions, companies, and public authorities, understanding these rules is an important part of AI ethics and regulatory compliance.

 

What are prohibited AI practices?

 

The EU AI Act classifies some AI systems as posing an unacceptable risk. Unlike high-risk AI systems, which may be used under strict conditions, prohibited AI systems are simply not allowed.

The reason is simple: some uses of AI can seriously harm people's rights, dignity, safety, or freedom.

Because, ethics is not only about following the law. It is about respecting people and preventing harm before it happens.

We bring you some examples of prohibited AI practices.

 

1. Manipulating People's Behaviour

 

AI must never manipulate people in ways they cannot easily recognise.

For example, an AI system should not use hidden techniques to pressure someone into making decisions that are against their own interests.

This is especially important when vulnerable people, such as children, older adults, or persons with disabilities, are involved.

 

2. Exploiting Vulnerable People

 

AI systems must not take advantage of people's age, disability, financial situation, or emotional condition. Responsible AI should support people—not exploit them.

Here is an example: an AI application should not encourage children to engage in dangerous behaviour or persuade vulnerable individuals to spend money they cannot afford.

 

3. Social Scoring

 

One of the best-known prohibited practices is social scoring. Many are still not familiar with this term.

Social scoring is the practice of using data analytics and artificial intelligence to monitor, evaluate, and assign a numerical score to individuals based on their behaviours, socioeconomic status, or personality. These scores can determine how they should be treated, and decide about a person's access to services, employment, public goods, or legal rights.

Denying someone access to services simply because an AI-generated score labels them as "undesirable" is considered incompatible with European values.

(Part 2 coming soon.)