EU Introduces World's First 'AI Regulation Law'... Accelerating Checks on US OpenAI and Google

Ban on Collecting Face and Biometric Data
Fines Up to 35 Million Euros for Violations
Stricter Than US AI Executive Order
Checks on US Big Tech Like OpenAI, MS, and Google

The European Union (EU) has agreed to introduce the world's first legislation to regulate artificial intelligence (AI) technology in order to secure leadership in AI regulation. This move is seen as an effort to curb American big tech companies accelerating AI development, such as OpenAI?which sparked the ChatGPT craze?Google, Meta, and Microsoft (MS), while buying time for European AI companies to catch up.

EU Agrees on World's First AI Regulation Law

On the 8th (local time), representatives from the EU Commission, the European Parliament, and the 27 EU member states agreed on the introduction of an AI regulation law that governs the use of AI technology and imposes fines on companies that violate the regulations.

The AI regulation law classifies AI technology risks into four categories based on criteria such as citizens' rights and threats to democracy. Technologies with the highest risk level, labeled as "unacceptable risk," such as facial recognition technology, are banned from mass collection of facial images and building databases (DBs). The collection of biometric information based on political beliefs, religion, sexual orientation, race, and other factors is also prohibited. However, exceptions are made to allow AI use for suspect searches in cases of serious crimes such as terrorism.

Additionally, images, videos, and texts generated by generative AI must be marked with a watermark indicating "AI-generated content."

Attention is also drawn to the introduction of guardrails (safety measures) for Artificial General Intelligence (AGI) that surpasses human intelligence. Companies operating AGI must summarize and report their model training methods and data. They are also obligated to establish policies that comply with EU copyright law. This applies to AI models with significant influence and systemic risks, such as OpenAI's large language model (LLM) "GPT-4," which will be subject to even stronger regulations.

The EU will impose fines of up to 35 million euros or 7% of global revenue on AI companies that violate the law. The legislation will take effect two years after approval by the European Parliament and Commission.

Buying Time for Domestic Companies to Develop AI Technology

This legislation comes amid active global discussions on AI regulation. Earlier, the United States signed an AI executive order at the end of October to prevent side effects caused by AI misuse. Both sides are rushing to introduce AI regulations and striving to establish global regulatory standards favorable to their domestic or regional companies.

While the U.S. AI executive order includes regulations such as mandatory safety assessments and AI content watermarking, along with support measures like easing visa requirements for foreign AI talent, the EU’s proposal focuses more heavily on regulation and imposes stricter regulatory standards. The regulatory content and penalties, including fines, are also more detailed.

Market analysts interpret the introduction of this regulation as Europe’s attempt to curb American AI big tech companies like OpenAI, Google, and Meta, given the absence of leading AI companies in Europe. In fact, the ultra-high-performance AI models subject to strong regulations classified by the EU as systemic risks are expected to include OpenAI’s GPT-4 and Google’s Gemini.

The U.S. daily The Washington Post noted, "While the U.S. Congress failed to devise measures for years, the EU introduced digital privacy regulations such as GDPR, which resulted in fines for Meta and Google, and Google had to delay the European launch of its generative AI chatbot Bard." It added, "The EU has agreed on groundbreaking AI legislation ahead of the U.S., and this AI regulation could pave the way for the EU’s regulatory proposal to become a global standard."

This content was produced with the assistance of AI translation services.