EU Approves World’s First Artificial Intelligence Laws
Last week, the European Union (EU) provided its final seal of approval for the world’s first comprehensive set of laws governing artificial intelligence (AI). The unanimous approval came from the ambassadors of the 27 EU countries, who rubber-stamped the political agreement reached in December. The aim of these regulations is to control the level of risk posed by AI systems.
Negotiations on the legal text began in June, but the process faced a fierce debate over the regulation of general-purpose AI, such as ChatGPT and Google’s Bard chatbot. EU policymakers eventually reached a political agreement on the main points of the AI Act, a bill that seeks to regulate AI systems based on their potential for harm.
The complexity of the law required over a month for technical refinement. On January 24th, the final version of the text was presented by the Belgian presidency of the Council of EU Ministers at a technical meeting. However, some member states raised concerns about not having enough time to fully analyze the text.
These concerns were eventually overcome with the adoption of the AI Act by the Committee of Permanent Representatives. But the approval from EU ambassadors was not guaranteed, as some powerful European countries resisted parts of the provisional deal until the last moments. France, Germany, and Italy, in particular, argued for a lighter regulatory regime for powerful AI models like Open AI’s GPT-4, which support general-purpose AI systems.
The European Parliament supported the implementation of hard rules for these models, as it deemed it unacceptable to exclude the most potent types of AI from regulation while burdening smaller actors with regulatory requirements. A compromise was eventually reached, leading to a tiered approach with horizontal transparency rules for all models and additional obligations for compelling models deemed to pose a systemic risk.
The EU Commission also approved the creation of AI factories to promote the adoption of generative AI in strategic sectors. This marks the first legally binding agreement on the regulation of AI worldwide.
While EU countries still have the opportunity to influence the implementation of the AI law through around 20 acts of secondary legislation, the AI Office, which will oversee AI models, is expected to receive significant staffing with seconded national experts.
The AI rulebook will be adopted by the European Parliament’s Internal Market and Civil Liberties Committees on February 13th, followed by a plenary vote provisionally scheduled for April 10-11. The final adoption will be complete with ministerial endorsement.
The approval of the world’s first AI laws by the EU signifies a major step toward regulating AI and ensuring its safe and responsible development and deployment for the benefit of the global community.
