In a groundbreaking development, the European Parliament and EU member states have reached a historic agreement on the world's first comprehensive laws to regulate artificial intelligence (AI). After an exhaustive 37-hour negotiation, Thierry Breton, the European Commissioner overseeing the legislation, hailed the deal as historic, emphasizing its significance in governing not only AI but also social media and search engines, including major players like X, TikTok, and Google. The negotiations, facilitated by Carme Artigas, Spain's secretary of state for AI, enjoyed support from France and Germany, despite reports suggesting tech companies in these countries sought a lighter regulatory touch to encourage innovation among smaller firms.
This landmark agreement places the EU at the forefront of global efforts to regulate AI, surpassing the US, China, and the UK. While specific details about the law remain sparse, it is expected to take effect no earlier than 2025. The negotiations witnessed clashes over foundation models and prolonged debates on AI-driven surveillance, with the European Parliament securing a ban on real-time surveillance and biometric technologies, including emotional recognition, with limited exceptions for unforeseen terrorist threats, victim searches, and serious crime prosecution.
MEP Brando Benefei, co-leading the parliament's negotiating team, highlighted the objective of delivering legislation that ensures AI development in Europe follows a human-centric approach, respecting fundamental rights and values. The agreement introduces a risk-based tiered system, with the highest level of regulation applied to machines posing the greatest risks to health, safety, and human rights. This includes the categorization of AI models based on "floating point operations per second" (Flops), with only one existing model, GPT4, falling into this highest-risk category. The lower tier of regulation imposes significant obligations on AI services, covering disclosure of data used for training, ranging from writing articles to diagnosing cancer.
Dragoș Tudorache, the Romanian MEP leading the European Parliament's four-year battle to regulate AI, emphasized the ban on AI technology determining or predetermining criminal behavior while acknowledging law enforcement's need for effective tools. The EU's commitment to avoiding past mistakes, where tech giants grew unchecked, reflects a determination to regulate content and behavior on digital platforms. This comprehensive regulation is expected to set an example for other governments, encouraging emulation of key aspects globally, as compliance with EU rules may extend to markets beyond the continent for AI companies.