STRASBOURG, France: The inaugural set of comprehensive regulations overseeing artificial intelligence has been established through a groundbreaking agreement following an extensive 37-hour negotiation between the European Parliament and EU member states.
Termed “historic” by Thierry Breton, the European Commissioner overseeing a spectrum of laws in Europe encompassing social media, search engines, and major entities like X, TikTok, and Google, this agreement marks a pivotal moment. Breton highlighted the dedication of 100 individuals who convened for nearly three days to finalize the accord, asserting that the sleepless hours invested were justified by the significance of this landmark deal. Carme Artigas, Spain’s secretary of state for AI, who facilitated the negotiations, noted France and Germany’s endorsement of the text, despite speculation that tech firms in these nations were advocating for a more lenient approach to nurture innovation among smaller companies.
Historic!
The EU becomes the very first continent to set clear rules for the use of AI 🇪🇺
The #AIAct is much more than a rulebook — it's a launchpad for EU startups and researchers to lead the global AI race.
The best is yet to come! 👍 pic.twitter.com/W9rths31MU
— Thierry Breton (@ThierryBreton) December 8, 2023
The EU’s agreement positions it as a frontrunner ahead of the US, China, and the UK in the race to oversee artificial intelligence, aiming to safeguard the public against potential threats to life posed by the rapidly evolving technology—a concern shared by many.
Few specifics were disclosed regarding the contents of the forthcoming law, slated to take effect no earlier than 2025. The negotiations between the European Parliament and EU member states were arduous, particularly concerning the development of foundational AI models for broad applications rather than specific ones. Additionally, prolonged discussions revolved around AI-driven surveillance, potentially utilized by law enforcement, employers, or retailers for real-time public monitoring and emotional analysis.
Among the outcomes, the European Parliament successfully advocated for restrictions on real-time surveillance and biometric technologies, including emotional recognition, albeit with three exceptions outlined by Breton. These exceptions would permit police use of such intrusive technologies in cases of unforeseen terrorist threats, victim searches, and the prosecution of severe crimes.
MEP Brando Benefei, co-leading the parliament’s negotiation team alongside Dragoș Tudorache, the Romanian MEP leading the four-year initiative to regulate AI, emphasized their achievement in securing a requirement for “independent authorities” to authorize “predictive policing.” This measure aims to prevent potential abuse by law enforcement and uphold the presumption of innocence in criminal matters.
Tudorache emphasized our objective: to craft legislation ensuring AI in Europe develops with a human-centric approach, upholding fundamental rights and values while fostering trust and awareness of maximizing the benefits of the ongoing AI revolution. He clarified during a press conference in Brussels that while supporting law enforcement tools to combat crime and ensure citizen safety, the ban specifically targets AI tech predicting criminal behavior.
The legislation’s core lies in a risk-tiered system, with the strictest rules for machines posing the highest risks to health, safety, and human rights. Originally, it encompassed systems with over 10,000 business users, but now defines high-risk based on computing transactions, measured in Flops, where currently only GPT4 falls within this new classification.
This law imposes significant obligations on AI services, mandating transparency about the data used to train machines for various tasks, from writing articles to diagnosing cancer. Tudorache proudly stated that the EU leads by example globally in establishing genuine AI regulations, steering its evolution toward a human-centric direction.
Referencing past oversights with tech giants like Facebook, Tudorache aims to avoid repeating history by ensuring stringent oversight. Experts like Anu Bradford suggest the EU’s robust regulation could serve as a model for other governments. They may not replicate every aspect but are likely to adopt many elements.
Furthermore, companies complying with EU rules may extend these obligations to markets beyond Europe for efficiency rather than training separate models for different regions, according to Bradford.
