The European Parliament has now also approved the new AI Act. This European AI Act aims to regulate the use of AI, in part by making the use of AI for malicious purposes a criminal offense. The so-called AI Act also establishes clear rules governing the use and application of artificial intelligence, the most important technology of our time!
What does the regulation of AI entail?
The AI Act aims to ensure that “certain AI applications that threaten citizens’ rights are banned.” The agreement among European Member States states that AI systems will be classified into different risk profiles. Systems falling into the highest risk category will be banned. However, even lower-risk systems must meet strict requirements to be approved by the Parliament. Consider, for example, algorithms used to screen resumes. These may be used, but a company must ensure that the system will not discriminate on the basis of race, age, gender, and so on.
What does AI regulation mean in practice?
Among other things, this means that companies are prohibited from developing biometric systems based on “sensitive characteristics.” They are also prohibited from scraping faces from other platforms for the purpose of creating a facial recognition database. AI that can manipulate human behavior or exploit people’s vulnerabilities is also prohibited. It is also prohibited to create systems that recognize emotions in schools or workplaces, assign social scores, or perform predictive policing. The police themselves will also be subject to clear guidelines regarding the use of artificial intelligence. They may not use biometric identification systems, except in special cases, but for such cases, they must submit a detailed explanation. Examples include cases such as a missing person or a terrorist threat. Finally, the AI Act stipulates that images and videos created using AI must also be clearly labeled as “AI content” so that users know they are not real. This is, of course, to combat deepfakes and similar content.
When will the new AI law take effect?
The new AI regulations will be implemented in phases, not all at once. This means it will take some time. The Parliament will not publish the final legislation until May, and it will take another six months after that for the ban on unacceptable AI systems to take effect. The rules for generative AI systems, such as ChatGPT and Midjourney, will not take effect until 2025. Finally, the last set of rules will take effect in 2026. Systems with a high-risk profile will then be required to undergo a human rights assessment to ensure that the system does not discriminate or apply biases. Companies that violate the AI law could face extremely high fines, which can reach up to 35 million euros.