2024-07-26 10:44:19
Millions in fines for violations
First EU AI law: Which technologies will soon be banned and why
Updated on 26.07.2024 – 09:58Reading time: 3 min.
The EU’s first AI law, which regulates the use of artificial intelligence, will soon come into force. What exactly it says and which AI will be banned in the future.
- The first EU AI law comes into force on August 1 and regulates the use of artificial intelligence.
- AI systems that are deemed to pose an unacceptable risk are prohibited, including those that manipulate human behavior.
- Violations of the AI Act can result in fines of up to 35 million euros or seven percent of annual global turnover.
Systems based on artificial intelligence (AI) are already used in many areas today. ChatGPT, Gemini and Perplexity are increasingly used in everyday life. But AI is also used in video surveillance, speech recognition and the analysis of financial data.
For a long time, there were no established regulations for this. On August 1, the European Union’s first AI law will come into force, which will clearly regulate the deployment and use of artificial intelligence in EU countries in the future.
According to the European Parliament, the law is based on a risk assessment of AI systems. Artificial intelligence that is classified as an unacceptable risk will be banned in the future. These include:
- AI systems that can manipulate human behavior and emotions, such as voice-controlled toys that encourage dangerous behavior in children.
- AI systems that classify people into groups based on criteria such as skin color, political and religious views, or sexual orientation.
- AI systems that create a “social credit” system. With this technology, which is common in China, the ruling Communist Party wants to “educate” people to desired behavior and political loyalty through mass surveillance.
The use of AI must not discriminate against people on the basis of their age, disability or financial situation. This is to ensure that AI systems are not misused and that the protection of fundamental rights is always guaranteed.
The use of AI in facial recognition is also generally prohibited. However, some exceptions can be permitted for law enforcement purposes. However, police and other security authorities will in future only be allowed to use AI-controlled facial recognition in public places if there is a court order.
If video footage is recorded, the technology can be used to search for convicted persons or suspects of serious crimes. If officers follow the video footage in real time, AI systems are also permitted to search for victims of human trafficking and sexual violence. They can also use facial recognition to avert “a concrete and acute terrorist threat.”
In addition to unacceptable risk, there are also high-risk AI systems. These systems should be constantly monitored and must meet a number of requirements in order to be used.
When it comes to the use of artificial intelligence in security agencies, the legislator generally sees a “high risk” of violations of fundamental rights. The same applies to applications in human resources management, for example. The so-called high-risk applications also include systems that can influence elections.
In these cases, a human should therefore have final control over the AI decisions. The law also requires technical documentation and a risk management system. Those affected should be able to file a complaint with the authorities against the use of AI.
The new law will introduce a transparency requirement. Artificially created or edited content, such as images, audio or videos, will have to be clearly marked as such in the future. Online platforms such as Facebook, Instagram or X and national authorities will have to check this. However, experts believe that monitoring is difficult.
Violations of the AI law can result in fines running into the millions. The EU Commission can demand payments of up to 35 million euros or seven percent of a company’s global annual turnover for the use of a prohibited technology. Anyone who violates other provisions of the law can expect fines of up to 15 million euros or three percent of annual turnover.
The law is intended to take effect gradually. From February 2, 2025, unacceptable technologies such as “social credit” systems will initially be banned in the member states. From August 2, 2025, the rules for generally usable AI systems, such as those for text or image creation, will apply. From August 2, 2026, all other provisions of the AI law will come into effect. The EU Commission expects an “AI pact” from hundreds of companies on voluntary self-control as early as this autumn.