Full implementation in August 2026, two years later
Differentiated regulation applied in four levels according to risk degree
The European Union (EU) took its first step on the 1st (local time) as the world's first 'Comprehensive Artificial Intelligence (AI) Regulation Act' came into effect.
The European Commission, the EU's executive body, stated on the day, "The AI Regulation Act is designed to ensure the reliability of AI developed and used within the EU, including safeguards to protect human fundamental rights."
The law applies differentiated regulations in four stages according to the level of risk that may arise when AI technology is used in specific products or fields. The higher the risk of negative impact, the stricter the regulation.
In particular, AI technologies used in healthcare, education, elections, and critical infrastructure are classified as high-risk. In such cases, AI use must be supervised by humans, and a risk management system must be established.
The use of AI technologies that may infringe on human fundamental rights is completely prohibited. This includes social scoring that assigns individual scores based on data related to personal characteristics and behavior, and the act of collecting facial images from the internet or CCTV to build databases. The use of real-time remote biometric identification systems by law enforcement agencies is also banned except in extremely exceptional cases.
General-purpose AI, such as Chat GPT, is subject to transparency obligations. This should be done by specifying the content used in the AI learning process, among other methods.
Starting from the day of enforcement, the regulations on prohibited technologies will apply after six months, and the obligations for general-purpose AI will apply after one year. Full implementation will begin in August 2026, two years later.
According to the Commission, providing false information about AI technology may result in fines of 1.5% of global annual revenue, and violations of mandatory regulations may incur fines of 3%. If the law is violated by using prohibited AI applications, fines can increase up to 7%.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
