English 
简体中文 
日本語 
香港中文 
Français 
Español 
Bahasa Indonesia 
한국어 
EU member states have finally reached an agreement on the landmark Artificial Intelligence (AI) Act, as reported by Politico on February 2. As the world’s first comprehensive law to regulate AI, the AI Act aims to facilitate the development and adoption of trustworthy and human-centric AI across the EU, while ensuring that high-risk AI systems are subject to strict rules and oversight. Under the agreement, the law will ban some applications of AI technology, such as biometric categorization systems that use sensitive characteristics, emotion recognition in the workplace and educational institutions, and AI systems that manipulate human behavior to circumvent their free will. In addition, it imposes stringent limits on high-risk use cases and sets obligations of transparency and stress-testing for the most advanced software models. The European Parliament is scheduled to vote on the rules in March or April, and they could formally become law as early as May.
The AI Act was initially proposed by the European Commission in 2021 and endorsed by EU countries last December. However, both Germany and France voiced opposition to the original rules package agreed in December, arguing that some provisions may impede innovation. Considering the rapid evolution of AI models like ChatGPT, the revised AI Act entails provisions that grant the Commission flexibility to overrule standards in response to technological advancements, thus ensuring the law’s adaptability to the changing AI landscape. Another pivotal element of the Act involves the establishment of an independent regulatory body, the AI Office, tasked with the oversight of General Purpose AI (GPAI) models, particularly those identified as posing systemic risks.
Sources:
https://www.dw.com/en/eu-states-approve-landmark-ai-rules/a-68158258
