The World’s First Artificial Intelligence Act

Under the new rules, certain AI applications are prohibited. This includes biometric categorization systems that rely on sensitive characteristics, as well as the indiscriminate scraping of facial images from the internet or CCTV footage for the purpose of creating facial recognition databases. The Act also forbids the use of emotion recognition in workplaces and schools, social scoring, predictive policing, and any AI technology that manipulates human behavior or exploits vulnerabilities.

Law enforcement agencies are barred from utilizing biometric identification systems, except in narrowly defined situations that are expressly listed. The real-time use of these systems is restricted to specific time frames and geographical areas, subject to judicial or administrative approval. Post-remote use of biometric identification is considered high-risk and requires judicial consent for cases linked to criminal offenses.

High-risk applications of AI technology, such as those in critical infrastructure, education, employment, healthcare, banking, law enforcement, migration, justice, and democratic processes, must undergo risk assessments, ensure transparency, accuracy, and human oversight, and allow citizens to lodge complaints and seek explanations for decisions affected by high-risk AI systems.

Furthermore, any artificial or manipulated images, audio, or video content, commonly known as «deepfakes,» must be clearly labeled as such.

To stimulate innovation and support small and medium-sized enterprises, the Act proposes the establishment of regulatory sandboxes and real-world testing opportunities at the national level. These resources will be made accessible to SMEs and startups to facilitate the development and training of new and innovative AI technologies before they are brought to market.