Europe Takes the Lead: Milestone Agreement on AI Regulation

In a historic move, Europe has achieved a provisional agreement on a groundbreaking set of rules governing the use of artificial intelligence (AI) within the European Union. This marks a significant stride toward becoming the first major global power to implement comprehensive laws regulating AI. The deal, reached after nearly 15 hours of negotiations following a marathon 24-hour debate, positions Europe as a pioneer in shaping global standards for AI.

The agreement places specific obligations on foundational models like ChatGPT and general-purpose AI systems (GPAI) to adhere to transparency requirements before entering the market. These requirements include the creation of technical documentation, compliance with EU copyright law, and the dissemination of detailed summaries regarding the content used for training.

Foundation models carrying high systemic risk must conduct model evaluations, assess and mitigate potential systemic risks, perform adversarial testing, report serious incidents to the European Commission, ensure cybersecurity measures, and disclose information about their energy efficiency. GPAIs with systemic risk may rely on codes of practice to comply with the new regulations.

Governments within the EU are restricted in their use of real-time biometric surveillance in public spaces, only permitting it in cases involving victims of certain crimes, the prevention of immediate threats such as terrorist attacks, and the search for individuals suspected of the most serious offenses.

The agreement explicitly prohibits cognitive behavioral manipulation, untargeted scraping of facial images from the internet or CCTV footage, social scoring, and biometric categorization systems used to deduce political, religious, philosophical beliefs, sexual orientation, and race.

Consumers gain the right to file complaints and receive meaningful explanations for AI-related decisions. Violations of the regulations may result in fines ranging from 7.5 million euros ($8.1 million) or 1.5% of turnover to 35 million euros or 7% of global turnover.

However, not all parties are celebrating the agreement. Business group DigitalEurope criticized the rules, seeing them as an additional burden for companies alongside recent legislation. DigitalEurope's Director General Cecilia Bonefeld-Dahl expressed concerns about the last-minute inclusion of regulations for foundation models, altering the initial risk-based approach.

Privacy rights group European Digital Rights also raised criticism, particularly regarding the legislation's acceptance of live public facial recognition across the EU. While the Parliament aimed to limit potential damages, the group's senior policy advisor, Ella Jakubowska, viewed the overall package on biometric surveillance and profiling as, at best, lukewarm.

The legislation is expected to take effect early next year, pending formal ratification from both sides, with implementation slated for two years afterward. Governments worldwide are grappling with the challenge of balancing the benefits of advanced AI technology with the necessity of establishing effective regulatory frameworks to manage its potential risks.