Soon the European AI Regulation will enter into force. In the earlier article on DigitaleOverheid.nl you could read what the AI Regulation will mean for your organization. In this article you can read more about the practices around Artificial Intelligence (AI) that will be banned from February 2025. The bans may affect your government organization.
There are many useful applications of AI, but this technology can also be misused for manipulation, exploitation and social control. The European AI regulation seeks to ban such practices to protect citizens.
Briefly, these are the prohibited AI practices:
Manipulative or misleading AI systems. While AI systems can help people make targeted and informed decisions, there is also a risk. AI techniques can influence people to make choices they would not normally make. For example, consider using audio, visual and video material that they cannot consciously perceive. Therefore, AI systems that use subconscious techniques and AI systems that purposefully manipulate or mislead are prohibited. AI systems that exploit vulnerabilities - such as age, disability or socioeconomic situation - to influence people are also prohibited.
AI systems for "social scoring. This means that AI systems judge people based on their social behavior or personal characteristics. This can cause problems, such as discrimination and exclusion of certain groups.
AI-based predicted crime risk assessments. AI can be used to assist in predicting possible criminal acts. However, it is important here that it is based on facts that are actually related to criminal activity. AI systems that predict crimes based solely on profiling or assessment of personality traits and characteristics are prohibited.
Biometric categorization. Systems that categorize individuals based on biometric data (such as facial images and fingerprints) to infer race, political views, union membership, religious or philosophical beliefs, sex life or sexual orientation are prohibited. This prohibition protects the privacy and fundamental rights of individuals. Exceptions only apply to lawful labeling or filtering of lawfully obtained biometric datasets, in the field of law enforcement.
The use of AI systems for real-time remote biometric identification in public places. This will be prohibited except when strictly necessary for specific and defined situations. Examples include searching for victims of kidnapping or missing persons, preventing imminent threats to people's lives, and tracking suspects of certain very serious crimes. Their deployment should be provided by national legislation with a legal basis detailing the conditions.
Other banned practices. Other AI practices banned from February 2025 include emotion recognition in the workplace and in education. Untargeted scrapping of facial images from the Internet or camera images (CCTV) to create or supplement facial recognition databases are also prohibited.
More information can be found in Article 5 of the AI Regulation on the website of EUR-Lex. The European Commission is still issuing guidelines on the practical implementation of the bans.
What does this mean for my organization?
It is important that organizations identify the prohibited AI systems they use so that they can phase them out. This must be done by February 2025. If this is not done in time, regulators can impose hefty fines.
On the Pleio area of the algorithm framework you will find a practical overview with the main existing standards. Here you will also find measures that can help with the deployment of algorithms and AI. Later, the requirements of the AI regulation will appear in the algorithm framework.
The AI regulation is a comprehensive law with many different provisions and considerations. What about what AI systems are allowed? You will read about the obligations and nuances involved soon on DigitaleOverheid.nl. This is the second article in a series we will publish in the coming weeks about the AI regulation. In it, we will discuss high-risk AI systems and oversight of the regulation, among other things.
Want to know more about the regulation timeline? Then go to the first article in this series.
