Since 2 February 2025 the EU AI Act (Regulation (EU) 2024/1689) bans a short list of AI practices across the EU, including in the Netherlands: manipulation, exploiting vulnerable people, social scoring, emotion recognition at work and in schools, certain biometric uses and untargeted scraping of facial images. There is no transition period, and the ban applies to businesses that develop AI as well as to businesses that only use it.
The ban concerns the use, not the technology. The same facial-recognition software can be lawful for access control to a server room and prohibited for other purposes. This page explains each prohibited practice, who enforces the ban in the Netherlands and what your business should do now.
Table of contents
Which AI practices are prohibited?
The list of prohibited practices in Article 5 of the AI Act contains eight categories. Most of them apply only when the practice causes, or is reasonably likely to cause, significant harm, or when it is aimed at certain purposes.
Manipulative or deceptive techniques
You may not use AI that deploys subliminal, purposefully manipulative or deceptive techniques to materially distort someone’s behaviour, in a way that causes or is likely to cause significant harm. Think of hidden nudging or dark patterns that push consumers into decisions they would not otherwise take. Ordinary, transparent personalisation of offers is not covered. Misleading commercial practices can also breach Dutch consumer law on unfair commercial practices.
Exploiting vulnerabilities
AI may not exploit the vulnerabilities of people due to their age, disability or a specific social or economic situation, with the aim or effect of materially distorting their behaviour and causing significant harm. Examples are systems that target children’s inexperience or push people in financial distress towards expensive credit.
Social scoring
Evaluating or classifying people over time based on their social behaviour or personal characteristics is banned where the score leads to detrimental treatment in an unrelated context, or treatment that is unjustified or disproportionate. The ban applies to public authorities and private businesses alike. Ordinary credit checks based on relevant financial data are not social scoring, but they may be high-risk.
Predicting crime from profiling
AI may not be used to assess the risk that a person will commit a criminal offence based solely on profiling or on personality traits. Systems that support a human assessment based on objective, verifiable facts linked to criminal activity are outside this ban.
Untargeted scraping of facial images
Creating or expanding facial recognition databases through untargeted scraping of facial images from the internet or from CCTV footage is prohibited. Buying access to such a database can make you a user of a prohibited practice.
Emotion recognition at work and in education
You may not use AI to infer the emotions of people in the workplace or in educational institutions, except for medical or safety reasons, such as detecting fatigue in a driver. This hits HR tools that analyse facial expressions or voice in job interviews, and call-centre software that scores employees’ emotions. If you plan any workplace monitoring, the works council’s consent under Article 27(1)(l) of the Works Councils Act (WOR) is also required.
Biometric categorisation of sensitive traits
AI may not categorise people on the basis of biometric data to deduce their race, political opinions, trade-union membership, religious or philosophical beliefs, sex life or sexual orientation. Labelling or filtering lawfully acquired biometric datasets in law enforcement is an exception.
Real-time remote biometric identification by the police
Real-time facial recognition in publicly accessible spaces for law enforcement is prohibited, except in narrowly defined cases such as searching for victims of abduction or trafficking, preventing an imminent threat to life or a terrorist attack, and locating suspects of serious crimes. Each use requires prior authorisation by a judicial or independent administrative authority. For businesses, the main point is that the exception is for law enforcement only.
Who has to comply?
Everyone in the chain: providers who develop or market an AI system, deployers who use it in their business, and importers and distributors. Buying a tool from a supplier does not shift your responsibility as the user.
The AI Act also applies to non-EU suppliers when the output of their system is used in the EU. Contracts with AI suppliers should therefore include a warranty that the system is not used for prohibited purposes and a right to terminate if it turns out to be.
Who enforces the ban in the Netherlands?
The Dutch act that designates the supervisors under the AI Act (Uitvoeringswet AI-verordening) is still a bill. The planned structure builds on existing sector supervisors, with the Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) for areas without a sector supervisor and the Dutch Authority for Digital Infrastructure (RDI) coordinating.
The absence of the implementing act does not make a prohibited practice lawful. Most prohibited practices also involve unlawful processing of personal data under the GDPR, and the AP can enforce the GDPR today. The European Commission published non-binding guidelines on the prohibited practices on 4 February 2025; the Court of Justice of the EU has the final word on interpretation.
What are the sanctions?
Breaching a prohibition carries the highest fine ceiling in the AI Act: a fixed maximum amount or a percentage of total worldwide annual turnover, whichever is higher. Lower ceilings apply to other breaches and to supplying incorrect information to an authority, and lower caps apply to small and medium-sized enterprises.
A person harmed by a prohibited system can also claim damages in the Dutch civil courts, for example on the basis of tort (onrechtmatige daad, Article 6:162 of the Dutch Civil Code). The proposed EU AI Liability Directive was withdrawn, so ordinary Dutch liability law applies.
What should you do now?
Start with an inventory of every AI system in use, including AI features built into software you already have, such as HR, marketing or security tools. Then test each use against the eight categories above.
- Record for each system its purpose, its input, its output and who it affects.
- Stop any prohibited use immediately and record when and how you stopped it.
- Ask suppliers in writing whether their system contains emotion recognition, biometric categorisation or scraped facial data.
- Keep your assessments, so that you can show a supervisor what you checked and when.
Does the AI literacy duty apply as well?
Yes. Since the same date, 2 February 2025, providers and deployers must take measures to ensure that staff who operate or use AI systems have a sufficient level of AI literacy. This applies to all AI systems, not only high-risk ones.
Training should fit the role: HR staff need to recognise a prohibited emotion-recognition feature in a recruitment tool, and procurement staff need to ask the right questions of suppliers. Keep records of who was trained, when and on what.
What should you check as a business using AI?
- Does any tool infer emotions of employees or job applicants?
- Does any system score customers on behaviour unrelated to the service you provide?
- Does any marketing tool target people on age, disability or financial distress in a way that could harm them?
- Do you use a facial recognition database, and do you know how it was built?
- Have you trained staff to recognise prohibited uses?
What should you check as an AI supplier?
- Can your product be configured for a prohibited use, and do your terms forbid that?
- Does your product infer emotions or sensitive traits from biometric data?
- Was any training data obtained through untargeted scraping of facial images?
- Do your instructions for use explain which uses are not allowed?
- Have you documented your assessment against the prohibited practices?
What can we do for you with prohibited AI practices?
Our IT lawyers help businesses check their AI use against the AI Act and the GDPR.
- We assess your AI inventory against the prohibited practices and the high-risk categories.
- We review supplier contracts and add warranties and termination rights for prohibited use.
- We draft your internal AI policy and AI literacy programme.
- We advise on the GDPR position of the same systems, where enforcement is already possible.
- We assist you in enquiries or investigations by the AP or a sector supervisor.
Summary
- Since 2 February 2025 the AI Act bans eight categories of AI practices, without a transition period.
- The ban applies to providers and to businesses that only use AI.
- Emotion recognition at work and in education is banned, except for medical or safety reasons.
- The Dutch implementing act is still a bill, but the AP can already act under the GDPR.
- Prohibited practices carry the highest fine ceiling in the AI Act; harmed persons can also claim damages.
Frequently asked questions
Since when are these AI practices prohibited?
Since 2 February 2025. The ban applies directly in all EU member states, including the Netherlands, without a transition period.
Is emotion recognition always prohibited?
No. It is prohibited in the workplace and in educational institutions, except for medical or safety reasons. Outside those settings it is not banned, but it is a high-risk use and the GDPR applies.
Is a credit check a form of prohibited social scoring?
Normally not. A credit check based on relevant financial data is not social scoring, but credit scoring of natural persons is a high-risk use under the AI Act.
Who can fine my business for using a prohibited AI system?
Once the Dutch implementing act is in force, the designated market surveillance authorities. Until then, the Dutch Data Protection Authority (AP) can act under the GDPR where personal data is processed.
