EU submits artificial intelligence regulation bill to Congress

Recruitment exams, entrance exams, and exam evaluations by autonomous vehicles and companies by the European Commission, an EU policy enforcement agency. A new law (European Approach for Artificial Intelligence) imposed on the use of artificial intelligence in terms of threats to human safety and fundamental rights in areas such as bank lending was submitted to Congress.

While information technology giants such as Google, Amazon, Facebook, and Microsoft are devoting considerable resources to AI development, Google CEO Sundar Pichai reported that AI should be regulated and that AI development that is contrary to ethics should be banned. In recent years, discussions on AI regulation have become active. In this trend, the European Commission said that AI is solving a variety of social problems, but this is limited to cases where AI is developed and used in a way that can obtain high quality and people’s trust, and that the AI system used in the EU is safe and without ethical prejudice. A new law was submitted to Congress to confirm that it was under human control.

This new law categorizes AI into four categories: unacceptable risk, high risk, limited risk, and minimal risk by field in which AI is used, and applies different restrictions for each category. Areas identified as unacceptable risks are considered public threats to EU citizens, such as a system similar to the Chinese social credit system, which ranks and manages all citizens, and an AI-assisted system that promotes child-risk behavior, and is subject to a total ban. High risk encompasses various fields such as infrastructure such as transport, which can endanger the lives and health of citizens, education, vocational training, examination and evaluation, and labor relations such as employment and worker management, and immigration, border management, etc. As well as after the start of operation, the evaluation continues carefully.

As a limited risk, one area imposes a duty of minimum transparency in systems that support informed decision-making, such as chatbots. The minimal risk seems to be not subject to this category limit in AI used in applications such as video games and spam filters.

According to reports based on previously leaked information, companies that violate this law are fined up to 20 million euros or up to 4% of sales, but in reality, the bill submitted to Congress is the world’s total annual revenue It is said to be fined up to 6% of the total. However, reports say that in reality the highest fines will seldom be imposed.

While some criticize that the bill submitted to Parliament has too many vague loopholes, voices are rising that it will overtake China because the cost of developing AI in the EU is either banned or technically impossible. The legislation is expected to take years to come into effect, as it needs to be approved by both the European Council and the European Parliament. Related information can be found here.